",
"text/plain": " Parameters\nnan_mode Min\neval_metric Logloss\niterations 1000\nsampling_frequency PerTree\nleaf_estimation_method Newton\ngrow_policy SymmetricTree\npenalties_coefficient 1\nboosting_type Plain\nmodel_shrink_mode Constant\nfeature_border_type GreedyLogSum\nbayesian_matrix_reg 0.10000000149011612\nl2_leaf_reg 3\nrandom_strength 1\nrsm 1\nboost_from_average False\nmodel_size_reg 0.5\nsubsample 0.800000011920929\nuse_best_model False\nclass_names [0, 1]\nrandom_seed 5221\ndepth 6\nposterior_sampling False\nborder_count 254\nclasses_count 0\nauto_class_weights None\nsparse_features_conflict_fraction 0\nleaf_estimation_backtracking AnyImprovement\nbest_model_min_trees 1\nmodel_shrink_rate 0\nmin_data_in_leaf 1\nloss_function Logloss\nlearning_rate 0.015974000096321106\nscore_function Cosine\ntask_type CPU\nleaf_estimation_iterations 10\nbootstrap_type MVS\nmax_leaves 64"
}
],
"_view_module": "@jupyter-widgets/output",
"_model_module_version": "1.0.0",
"_view_count": null,
"_view_module_version": "1.0.0",
"layout": "IPY_MODEL_b995a9f6bf614675ac2153fc65daa3de",
"_model_module": "@jupyter-widgets/output"
}
},
"497321ff3b6849b19b7ab938133316ad": {
"model_module": "@jupyter-widgets/controls",
"model_name": "ToggleButtonsStyleModel",
"state": {
"_view_name": "StyleView",
"button_width": "",
"_model_name": "ToggleButtonsStyleModel",
"description_width": "",
"_view_module": "@jupyter-widgets/base",
"_model_module_version": "1.5.0",
"_view_count": null,
"font_weight": "",
"_view_module_version": "1.2.0",
"_model_module": "@jupyter-widgets/controls"
}
},
"cc0fbcf2f77846f4bd1266103d07a1f7": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"state": {
"_view_name": "LayoutView",
"grid_template_rows": null,
"right": null,
"justify_content": null,
"_view_module": "@jupyter-widgets/base",
"overflow": null,
"_model_module_version": "1.2.0",
"_view_count": null,
"flex_flow": null,
"width": null,
"min_width": null,
"border": null,
"align_items": null,
"bottom": null,
"_model_module": "@jupyter-widgets/base",
"top": null,
"grid_column": null,
"overflow_y": null,
"overflow_x": null,
"grid_auto_flow": null,
"grid_area": null,
"grid_template_columns": null,
"flex": null,
"_model_name": "LayoutModel",
"justify_items": null,
"grid_row": null,
"max_height": null,
"align_content": null,
"visibility": null,
"align_self": null,
"height": null,
"min_height": null,
"padding": null,
"grid_auto_rows": null,
"grid_gap": null,
"max_width": null,
"order": null,
"_view_module_version": "1.2.0",
"grid_template_areas": null,
"object_position": null,
"object_fit": null,
"grid_auto_columns": null,
"margin": null,
"display": null,
"left": null
}
},
"b995a9f6bf614675ac2153fc65daa3de": {
"model_module": "@jupyter-widgets/base",
"model_name": "LayoutModel",
"state": {
"_view_name": "LayoutView",
"grid_template_rows": null,
"right": null,
"justify_content": null,
"_view_module": "@jupyter-widgets/base",
"overflow": null,
"_model_module_version": "1.2.0",
"_view_count": null,
"flex_flow": null,
"width": null,
"min_width": null,
"border": null,
"align_items": null,
"bottom": null,
"_model_module": "@jupyter-widgets/base",
"top": null,
"grid_column": null,
"overflow_y": null,
"overflow_x": null,
"grid_auto_flow": null,
"grid_area": null,
"grid_template_columns": null,
"flex": null,
"_model_name": "LayoutModel",
"justify_items": null,
"grid_row": null,
"max_height": null,
"align_content": null,
"visibility": null,
"align_self": null,
"height": null,
"min_height": null,
"padding": null,
"grid_auto_rows": null,
"grid_gap": null,
"max_width": null,
"order": null,
"_view_module_version": "1.2.0",
"grid_template_areas": null,
"object_position": null,
"object_fit": null,
"grid_auto_columns": null,
"margin": null,
"display": null,
"left": null
}
}
}
}
},
"cells": [
{
"cell_type": "markdown",
"metadata": {
"id": "TRQ197E7_GeW"
},
"source": [
"**A Study on the Correlation of Weather Station Data and the Occurence of Wet Avalanches**"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "iWjUyErHoy1p"
},
"source": [
"# Contribution"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "BGCyS_6gpIDA"
},
"source": [
"\n",
"\n",
"Mahad:\n",
"\n",
"\n",
"> Mahad handled all of the practical analysis done on the data. Mahad also performed most of the data organization and machine learning done on the data.\n",
"\n",
"\n",
"Grace: \n",
"\n",
"\n",
"> Grace handled all of the linear and logistic regression performed on the data. Grace did assist in organizing and cleaning up the data for her own practice; however, the code used for these tasks in this notebook is Mahad's. Grace organized and prepared the notebooks for submission. Grace wrote the abstract, introduction, and other pieces of information throughout the notebook. Lastly, she helped to research details about wet avalanches as well as different methods for analyzing the data.\n",
"\n",
"\n",
"\n",
"Daniela:\n",
"\n",
"\n",
"> Daniela communicated with sponsors, professor, and team mates. She set up a meeting with the sponsors and kept track of all emails and data. Daniela also recorded all information and next steps during presentations. She helped with researching Pycaret and defaults along with researching methods that showed the most accuracy.Worked on the testing and training portion. Looked at if there were changes between 60:40, 70:30, and 80:20.\n",
"\n",
"Required Files:\n",
"\n",
"The following xlsx files contain the data needed to run the notebook and are available in the CAIC drive folder. They are read in under 'Data Import and Pre-Processing'\n",
"* All Wet Avalanches 2014-Present.xlsx\n",
"* OriginalWeatherData.xlsx\n",
"\n",
"**It is important to note that all of the Machine Learning Code underneath the sub section 'Prediction Model' takes at least a minute, likely more, to run for each box of code**\n",
"\n"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "O2TdhRU4qydE"
},
"source": [
"# Abstract"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "mrBEngaah-RG"
},
"source": [
"This report organizes and provides analysis for data provided by the Colorado Avalanche Information Center, regarding wet avalanches. Our goal was to clean and subsequently analyze the wet avalanche data recorded by CAIC as well as the corresponding weather station data in order to find any correlation between the various data points. Overall, this notebook provides a template for organizing, analyzing and predicting wet avalanches; however, the predictions could be more precise with more locations. In the future, we hope that this information can be used to study the causes of wet avalanches in order to predict when they occur. "
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "yul-lpNSg75v"
},
"source": [
"# Introduction"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "8inJZ_uC_T6O"
},
"source": [
"\n",
"The Colorado Avalanche Information Center is a program within the Colorado Department of Natural Resources that is committed to educating the public on avalanche safety as well as forecasting avalanche conditions. Currently, the CAIC employs a system that is titled '[The Avalanche Problem](https://www.avalanche.state.co.us/forecasts/help/avalanche-problems/)', which uses four characteristics (avalanche character or type, location, likelihood, and size) to determine the avalanche hazard rating throughout the mountains each day. \n",
"\n",
"Our research focuses on two of these avalanche characters, Wet Slab and Wet Loose, which we have combined to the single term 'wet avalanches' for the sake of the study. There is little research surrounding wet avalanches and their causes, so our goal was to take the weather station data and wet avalanche data recorded by the CAIC and study their correlation. We hope that the research that we have done can be used as a first-step in identifying where and why wet avalanches form. \n",
"\n",
"At the start of our research, we relied primarily on graphics in order to visualize the status of various weather varibles alongside the occurence of wet avalanches. Time series and distribution plots were used for this portion of the research. Linear regression was used to help identify how each weather variable directly correlates to avalanche occurence, and machine learning tools were implemented to start generating predictive modeling options for future use. "
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "H5o-hqqWAB1o"
},
"source": [
"# Methods"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "W6HBR_IhAmd6"
},
"source": [
"There are two key pieces of data that we are using in this project, wet avalanche data and weather station data. The wet avalanche data was given to us by the CAIC team in the form of a 664 KB .csv file. We then turned it into a .xlsx file (All Wet Avalanches 2014-Present.xlsx); however, the .csv form of the file is still available in the CAIC drive folder. The SNOTEL weather station data was downloaded from the [National Resources Conservation Center](https://www.wcc.nrcs.usda.gov/snow/snotel-data.html) website. There are 13 SNOTEL stations that were used in this project; so, the data from each station were combined into one .xlsx file (OriginalWeatherData.xlsx) and is 6 MB. The wet avalance data is both categorical and numerical. A description of each column and its units is listed below:"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "EX9nC9iuPfw8"
},
"source": [
"Columns and units for 'All Wet Avalanches 2014-Present.xlsx':\n",
"\n",
"1. id = id \n",
"2. obs_id = Observation id \n",
"3. avi_hw_op_bc = Was the avalanche in a highway, within an operation (ski area) or backcountry\n",
"4. avi_hw_zone_id = If highway what is the pass id (-1 = not highway)\n",
"5. avi_path = Avalanche path name if known. These are mostly highway avalanche paths where the name is known.\n",
"6. avi_op_name = name of operation if within an operating boundary\n",
"7. avi_loc = general area of avalanche from a drop down list\n",
"8. avi_bc_zone_id = if avalanche is a backcountry avalanche which CAIC zone is it in\n",
"9. avi_mark = Location within a backcountry zone if known\n",
"10. avi_number = number of avalanches reported at that place and time\n",
"11. avi_type = type of avalanche (WL = wet loose, WS = wet slab)\n",
"12. avi_aspect = Compas aspect if known\n",
"13. avi_elev = elevation compared to treeline can be (>TL = above treeline, TL = at treeline, \n",
"\n",
"
Precipitation Accumulation (in) Start of Day Values
\n",
"
Air Temperature Maximum (degF)
\n",
"
Air Temperature Minimum (degF)
\n",
"
Air Temperature Average (degF)
\n",
"
Precipitation Increment (in)
\n",
"
\n",
" \n",
" \n",
"
\n",
"
0
\n",
"
Berthoud Pass
\n",
"
1978-10-01
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.0
\n",
"
\n",
"
\n",
"
1
\n",
"
Berthoud Pass
\n",
"
1978-10-02
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.0
\n",
"
\n",
"
\n",
"
2
\n",
"
Berthoud Pass
\n",
"
1978-10-03
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.0
\n",
"
\n",
"
\n",
"
3
\n",
"
Berthoud Pass
\n",
"
1978-10-04
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.2
\n",
"
\n",
"
\n",
"
4
\n",
"
Berthoud Pass
\n",
"
1978-10-05
\n",
"
0.0
\n",
"
0.2
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.0
\n",
"
\n",
"
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
\n",
"
\n",
"
174871
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-03
\n",
"
0.0
\n",
"
NaN
\n",
"
60.0
\n",
"
41.0
\n",
"
49.0
\n",
"
NaN
\n",
"
\n",
"
\n",
"
174872
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-04
\n",
"
0.0
\n",
"
0.0
\n",
"
59.0
\n",
"
38.0
\n",
"
48.0
\n",
"
0.0
\n",
"
\n",
"
\n",
"
174873
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-05
\n",
"
0.0
\n",
"
0.0
\n",
"
62.0
\n",
"
42.0
\n",
"
51.0
\n",
"
0.0
\n",
"
\n",
"
\n",
"
174874
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-06
\n",
"
0.0
\n",
"
0.0
\n",
"
62.0
\n",
"
44.0
\n",
"
52.0
\n",
"
0.0
\n",
"
\n",
"
\n",
"
174875
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-07
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
\n",
" \n",
"
\n",
"
174876 rows × 8 columns
\n",
"
"
],
"text/plain": [
" Location Date \\\n",
"0 Berthoud Pass 1978-10-01 \n",
"1 Berthoud Pass 1978-10-02 \n",
"2 Berthoud Pass 1978-10-03 \n",
"3 Berthoud Pass 1978-10-04 \n",
"4 Berthoud Pass 1978-10-05 \n",
"... ... ... \n",
"174871 Wolf Creek pass Snotel 2020-10-03 \n",
"174872 Wolf Creek pass Snotel 2020-10-04 \n",
"174873 Wolf Creek pass Snotel 2020-10-05 \n",
"174874 Wolf Creek pass Snotel 2020-10-06 \n",
"174875 Wolf Creek pass Snotel 2020-10-07 \n",
"\n",
" Snow Water Equivalent (in) Start of Day Values \\\n",
"0 0.0 \n",
"1 0.0 \n",
"2 0.0 \n",
"3 0.0 \n",
"4 0.0 \n",
"... ... \n",
"174871 0.0 \n",
"174872 0.0 \n",
"174873 0.0 \n",
"174874 0.0 \n",
"174875 0.0 \n",
"\n",
" Precipitation Accumulation (in) Start of Day Values \\\n",
"0 0.0 \n",
"1 0.0 \n",
"2 0.0 \n",
"3 0.0 \n",
"4 0.2 \n",
"... ... \n",
"174871 NaN \n",
"174872 0.0 \n",
"174873 0.0 \n",
"174874 0.0 \n",
"174875 0.0 \n",
"\n",
" Air Temperature Maximum (degF) Air Temperature Minimum (degF) \\\n",
"0 NaN NaN \n",
"1 NaN NaN \n",
"2 NaN NaN \n",
"3 NaN NaN \n",
"4 NaN NaN \n",
"... ... ... \n",
"174871 60.0 41.0 \n",
"174872 59.0 38.0 \n",
"174873 62.0 42.0 \n",
"174874 62.0 44.0 \n",
"174875 NaN NaN \n",
"\n",
" Air Temperature Average (degF) Precipitation Increment (in) \n",
"0 NaN 0.0 \n",
"1 NaN 0.0 \n",
"2 NaN 0.0 \n",
"3 NaN 0.2 \n",
"4 NaN 0.0 \n",
"... ... ... \n",
"174871 49.0 NaN \n",
"174872 48.0 0.0 \n",
"174873 51.0 0.0 \n",
"174874 52.0 0.0 \n",
"174875 NaN NaN \n",
"\n",
"[174876 rows x 8 columns]"
]
},
"metadata": {
"tags": []
},
"execution_count": 12
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "On8sB-1TsGa_"
},
"source": [
"We can see now that all the stations have their data properly labeled into a unique dataset."
]
},
{
"cell_type": "code",
"metadata": {
"id": "MtsrLrSgFPK_",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "0e288c6f-a257-4a13-9062-02b6e6ff50ed"
},
"source": [
"# Get information about the dataframe\n",
"\n",
"CombinedDataWeather.info()"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"\n",
"RangeIndex: 174876 entries, 0 to 174875\n",
"Data columns (total 8 columns):\n",
" # Column Non-Null Count Dtype \n",
"--- ------ -------------- ----- \n",
" 0 Location 174876 non-null object \n",
" 1 Date 174876 non-null object \n",
" 2 Snow Water Equivalent (in) Start of Day Values 173843 non-null float64\n",
" 3 Precipitation Accumulation (in) Start of Day Values 173798 non-null float64\n",
" 4 Air Temperature Maximum (degF) 159835 non-null float64\n",
" 5 Air Temperature Minimum (degF) 159788 non-null float64\n",
" 6 Air Temperature Average (degF) 160247 non-null float64\n",
" 7 Precipitation Increment (in) 173777 non-null float64\n",
"dtypes: float64(6), object(2)\n",
"memory usage: 10.7+ MB\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "ZeW9LMN0sODl"
},
"source": [
"As it was done with the avalanches data, we also need to change the date variable to datetime format."
]
},
{
"cell_type": "code",
"metadata": {
"id": "KrkAz30GFeBg",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "94378459-df51-456f-cbe6-da25ed9c1f9d"
},
"source": [
"# Adjust variables types\n",
"\n",
"CombinedDataWeather['Date'] = pd.to_datetime(CombinedDataWeather['Date'])\n",
"\n",
"CombinedDataWeather.info()"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"\n",
"RangeIndex: 174876 entries, 0 to 174875\n",
"Data columns (total 8 columns):\n",
" # Column Non-Null Count Dtype \n",
"--- ------ -------------- ----- \n",
" 0 Location 174876 non-null object \n",
" 1 Date 174876 non-null datetime64[ns]\n",
" 2 Snow Water Equivalent (in) Start of Day Values 173843 non-null float64 \n",
" 3 Precipitation Accumulation (in) Start of Day Values 173798 non-null float64 \n",
" 4 Air Temperature Maximum (degF) 159835 non-null float64 \n",
" 5 Air Temperature Minimum (degF) 159788 non-null float64 \n",
" 6 Air Temperature Average (degF) 160247 non-null float64 \n",
" 7 Precipitation Increment (in) 173777 non-null float64 \n",
"dtypes: datetime64[ns](1), float64(6), object(1)\n",
"memory usage: 10.7+ MB\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "K3JXXea5OGNl",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 793
},
"outputId": "0082981b-8353-4aab-cc49-c602a765eb2a"
},
"source": [
"CombinedDataWeather"
],
"execution_count": null,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/html": [
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
Location
\n",
"
Date
\n",
"
Snow Water Equivalent (in) Start of Day Values
\n",
"
Precipitation Accumulation (in) Start of Day Values
\n",
"
Air Temperature Maximum (degF)
\n",
"
Air Temperature Minimum (degF)
\n",
"
Air Temperature Average (degF)
\n",
"
Precipitation Increment (in)
\n",
"
\n",
" \n",
" \n",
"
\n",
"
0
\n",
"
Berthoud Pass
\n",
"
1978-10-01
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.0
\n",
"
\n",
"
\n",
"
1
\n",
"
Berthoud Pass
\n",
"
1978-10-02
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.0
\n",
"
\n",
"
\n",
"
2
\n",
"
Berthoud Pass
\n",
"
1978-10-03
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.0
\n",
"
\n",
"
\n",
"
3
\n",
"
Berthoud Pass
\n",
"
1978-10-04
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.2
\n",
"
\n",
"
\n",
"
4
\n",
"
Berthoud Pass
\n",
"
1978-10-05
\n",
"
0.0
\n",
"
0.2
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.0
\n",
"
\n",
"
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
\n",
"
\n",
"
174871
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-03
\n",
"
0.0
\n",
"
NaN
\n",
"
60.0
\n",
"
41.0
\n",
"
49.0
\n",
"
NaN
\n",
"
\n",
"
\n",
"
174872
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-04
\n",
"
0.0
\n",
"
0.0
\n",
"
59.0
\n",
"
38.0
\n",
"
48.0
\n",
"
0.0
\n",
"
\n",
"
\n",
"
174873
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-05
\n",
"
0.0
\n",
"
0.0
\n",
"
62.0
\n",
"
42.0
\n",
"
51.0
\n",
"
0.0
\n",
"
\n",
"
\n",
"
174874
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-06
\n",
"
0.0
\n",
"
0.0
\n",
"
62.0
\n",
"
44.0
\n",
"
52.0
\n",
"
0.0
\n",
"
\n",
"
\n",
"
174875
\n",
"
Wolf Creek pass Snotel
\n",
"
2020-10-07
\n",
"
0.0
\n",
"
0.0
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
\n",
" \n",
"
\n",
"
174876 rows × 8 columns
\n",
"
"
],
"text/plain": [
" Location Date \\\n",
"0 Berthoud Pass 1978-10-01 \n",
"1 Berthoud Pass 1978-10-02 \n",
"2 Berthoud Pass 1978-10-03 \n",
"3 Berthoud Pass 1978-10-04 \n",
"4 Berthoud Pass 1978-10-05 \n",
"... ... ... \n",
"174871 Wolf Creek pass Snotel 2020-10-03 \n",
"174872 Wolf Creek pass Snotel 2020-10-04 \n",
"174873 Wolf Creek pass Snotel 2020-10-05 \n",
"174874 Wolf Creek pass Snotel 2020-10-06 \n",
"174875 Wolf Creek pass Snotel 2020-10-07 \n",
"\n",
" Snow Water Equivalent (in) Start of Day Values \\\n",
"0 0.0 \n",
"1 0.0 \n",
"2 0.0 \n",
"3 0.0 \n",
"4 0.0 \n",
"... ... \n",
"174871 0.0 \n",
"174872 0.0 \n",
"174873 0.0 \n",
"174874 0.0 \n",
"174875 0.0 \n",
"\n",
" Precipitation Accumulation (in) Start of Day Values \\\n",
"0 0.0 \n",
"1 0.0 \n",
"2 0.0 \n",
"3 0.0 \n",
"4 0.2 \n",
"... ... \n",
"174871 NaN \n",
"174872 0.0 \n",
"174873 0.0 \n",
"174874 0.0 \n",
"174875 0.0 \n",
"\n",
" Air Temperature Maximum (degF) Air Temperature Minimum (degF) \\\n",
"0 NaN NaN \n",
"1 NaN NaN \n",
"2 NaN NaN \n",
"3 NaN NaN \n",
"4 NaN NaN \n",
"... ... ... \n",
"174871 60.0 41.0 \n",
"174872 59.0 38.0 \n",
"174873 62.0 42.0 \n",
"174874 62.0 44.0 \n",
"174875 NaN NaN \n",
"\n",
" Air Temperature Average (degF) Precipitation Increment (in) \n",
"0 NaN 0.0 \n",
"1 NaN 0.0 \n",
"2 NaN 0.0 \n",
"3 NaN 0.2 \n",
"4 NaN 0.0 \n",
"... ... ... \n",
"174871 49.0 NaN \n",
"174872 48.0 0.0 \n",
"174873 51.0 0.0 \n",
"174874 52.0 0.0 \n",
"174875 NaN NaN \n",
"\n",
"[174876 rows x 8 columns]"
]
},
"metadata": {
"tags": []
},
"execution_count": 15
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "CytvevGSt8aH"
},
"source": [
"In this notebook we will average the weather data by date."
]
},
{
"cell_type": "code",
"metadata": {
"id": "sNHBcIGKJ5-V",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 606
},
"outputId": "39a3b9a2-a0bd-44af-dec6-972d28577db1"
},
"source": [
"AveragedDataWeather = CombinedDataWeather.groupby(['Date']).mean().reset_index()\n",
"\n",
"AveragedDataWeather"
],
"execution_count": null,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/html": [
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
Date
\n",
"
Snow Water Equivalent (in) Start of Day Values
\n",
"
Precipitation Accumulation (in) Start of Day Values
\n",
"
Air Temperature Maximum (degF)
\n",
"
Air Temperature Minimum (degF)
\n",
"
Air Temperature Average (degF)
\n",
"
Precipitation Increment (in)
\n",
"
\n",
" \n",
" \n",
"
\n",
"
0
\n",
"
1978-10-01
\n",
"
0.000000
\n",
"
0.000000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
1
\n",
"
1978-10-02
\n",
"
0.000000
\n",
"
0.000000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
2
\n",
"
1978-10-03
\n",
"
0.000000
\n",
"
0.000000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
3
\n",
"
1978-10-04
\n",
"
0.000000
\n",
"
0.000000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.100000
\n",
"
\n",
"
\n",
"
4
\n",
"
1978-10-05
\n",
"
0.000000
\n",
"
0.100000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
\n",
"
\n",
"
15343
\n",
"
2020-10-03
\n",
"
0.023077
\n",
"
0.008333
\n",
"
58.000000
\n",
"
33.692308
\n",
"
43.846154
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
15344
\n",
"
2020-10-04
\n",
"
0.030769
\n",
"
0.007692
\n",
"
61.000000
\n",
"
34.000000
\n",
"
45.846154
\n",
"
0.008333
\n",
"
\n",
"
\n",
"
15345
\n",
"
2020-10-05
\n",
"
0.030769
\n",
"
0.016667
\n",
"
60.923077
\n",
"
35.769231
\n",
"
46.307692
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
15346
\n",
"
2020-10-06
\n",
"
0.107692
\n",
"
0.008333
\n",
"
64.615385
\n",
"
35.846154
\n",
"
47.923077
\n",
"
0.009091
\n",
"
\n",
"
\n",
"
15347
\n",
"
2020-10-07
\n",
"
0.092308
\n",
"
0.018182
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
\n",
" \n",
"
\n",
"
15348 rows × 7 columns
\n",
"
"
],
"text/plain": [
" Date Snow Water Equivalent (in) Start of Day Values \\\n",
"0 1978-10-01 0.000000 \n",
"1 1978-10-02 0.000000 \n",
"2 1978-10-03 0.000000 \n",
"3 1978-10-04 0.000000 \n",
"4 1978-10-05 0.000000 \n",
"... ... ... \n",
"15343 2020-10-03 0.023077 \n",
"15344 2020-10-04 0.030769 \n",
"15345 2020-10-05 0.030769 \n",
"15346 2020-10-06 0.107692 \n",
"15347 2020-10-07 0.092308 \n",
"\n",
" Precipitation Accumulation (in) Start of Day Values \\\n",
"0 0.000000 \n",
"1 0.000000 \n",
"2 0.000000 \n",
"3 0.000000 \n",
"4 0.100000 \n",
"... ... \n",
"15343 0.008333 \n",
"15344 0.007692 \n",
"15345 0.016667 \n",
"15346 0.008333 \n",
"15347 0.018182 \n",
"\n",
" Air Temperature Maximum (degF) Air Temperature Minimum (degF) \\\n",
"0 NaN NaN \n",
"1 NaN NaN \n",
"2 NaN NaN \n",
"3 NaN NaN \n",
"4 NaN NaN \n",
"... ... ... \n",
"15343 58.000000 33.692308 \n",
"15344 61.000000 34.000000 \n",
"15345 60.923077 35.769231 \n",
"15346 64.615385 35.846154 \n",
"15347 NaN NaN \n",
"\n",
" Air Temperature Average (degF) Precipitation Increment (in) \n",
"0 NaN 0.000000 \n",
"1 NaN 0.000000 \n",
"2 NaN 0.000000 \n",
"3 NaN 0.100000 \n",
"4 NaN 0.000000 \n",
"... ... ... \n",
"15343 43.846154 0.000000 \n",
"15344 45.846154 0.008333 \n",
"15345 46.307692 0.000000 \n",
"15346 47.923077 0.009091 \n",
"15347 NaN NaN \n",
"\n",
"[15348 rows x 7 columns]"
]
},
"metadata": {
"tags": []
},
"execution_count": 16
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "ta1onB3qKZ8v"
},
"source": [
"The grouping was successfully made so now we will merge that information in the avalanches dataset by using Date as the primary key to connect the datasets."
]
},
{
"cell_type": "code",
"metadata": {
"id": "I4OzjxAiOVIs",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 677
},
"outputId": "d52209a6-3242-412a-f304-5e37fafe5950"
},
"source": [
"# Merge Avalanches DataFrame with Weather DataFrame\n",
"\n",
"# FullData = pd.merge(CombinedDataWeather, DataAvalanches[['avi_mark', 'avi_date', 'Avalanche', 'avi_number', 'avi_type']], how='left', left_on=['Location', 'Date'], right_on=['avi_mark', 'avi_date'])\n",
"\n",
"FullData = pd.merge(DataAvalanches[['avi_mark', 'avi_date', 'Avalanche', 'avi_number', 'avi_type']], AveragedDataWeather, how='outer', left_on='avi_date', right_on='Date')\n",
"\n",
"FullData.sort_values(by='Date', inplace=True)\n",
"\n",
"FullData"
],
"execution_count": null,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/html": [
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
avi_mark
\n",
"
avi_date
\n",
"
Avalanche
\n",
"
avi_number
\n",
"
avi_type
\n",
"
Date
\n",
"
Snow Water Equivalent (in) Start of Day Values
\n",
"
Precipitation Accumulation (in) Start of Day Values
\n",
"
Air Temperature Maximum (degF)
\n",
"
Air Temperature Minimum (degF)
\n",
"
Air Temperature Average (degF)
\n",
"
Precipitation Increment (in)
\n",
"
\n",
" \n",
" \n",
"
\n",
"
1499
\n",
"
NaN
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
1978-10-01
\n",
"
0.000000
\n",
"
0.000000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
1500
\n",
"
NaN
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
1978-10-02
\n",
"
0.000000
\n",
"
0.000000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
1501
\n",
"
NaN
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
1978-10-03
\n",
"
0.000000
\n",
"
0.000000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
1502
\n",
"
NaN
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
1978-10-04
\n",
"
0.000000
\n",
"
0.000000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.100000
\n",
"
\n",
"
\n",
"
1503
\n",
"
NaN
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
1978-10-05
\n",
"
0.000000
\n",
"
0.100000
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
0.000000
\n",
"
\n",
"
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
...
\n",
"
\n",
"
\n",
"
16369
\n",
"
NaN
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
2020-10-06
\n",
"
0.107692
\n",
"
0.008333
\n",
"
64.615385
\n",
"
35.846154
\n",
"
47.923077
\n",
"
0.009091
\n",
"
\n",
"
\n",
"
16370
\n",
"
NaN
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
2020-10-07
\n",
"
0.092308
\n",
"
0.018182
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
\n",
"
\n",
"
985
\n",
"
Hoosier Pass
\n",
"
NaT
\n",
"
Yes
\n",
"
1.0
\n",
"
WS
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
\n",
"
\n",
"
1001
\n",
"
Independence Pass-East side
\n",
"
1970-01-01
\n",
"
Yes
\n",
"
1.0
\n",
"
WL
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
\n",
"
\n",
"
1002
\n",
"
NaN
\n",
"
1970-01-01
\n",
"
Yes
\n",
"
1.0
\n",
"
WS
\n",
"
NaT
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
NaN
\n",
"
\n",
" \n",
"
\n",
"
16371 rows × 12 columns
\n",
"
"
],
"text/plain": [
" avi_mark avi_date Avalanche avi_number avi_type \\\n",
"1499 NaN NaT NaN NaN NaN \n",
"1500 NaN NaT NaN NaN NaN \n",
"1501 NaN NaT NaN NaN NaN \n",
"1502 NaN NaT NaN NaN NaN \n",
"1503 NaN NaT NaN NaN NaN \n",
"... ... ... ... ... ... \n",
"16369 NaN NaT NaN NaN NaN \n",
"16370 NaN NaT NaN NaN NaN \n",
"985 Hoosier Pass NaT Yes 1.0 WS \n",
"1001 Independence Pass-East side 1970-01-01 Yes 1.0 WL \n",
"1002 NaN 1970-01-01 Yes 1.0 WS \n",
"\n",
" Date Snow Water Equivalent (in) Start of Day Values \\\n",
"1499 1978-10-01 0.000000 \n",
"1500 1978-10-02 0.000000 \n",
"1501 1978-10-03 0.000000 \n",
"1502 1978-10-04 0.000000 \n",
"1503 1978-10-05 0.000000 \n",
"... ... ... \n",
"16369 2020-10-06 0.107692 \n",
"16370 2020-10-07 0.092308 \n",
"985 NaT NaN \n",
"1001 NaT NaN \n",
"1002 NaT NaN \n",
"\n",
" Precipitation Accumulation (in) Start of Day Values \\\n",
"1499 0.000000 \n",
"1500 0.000000 \n",
"1501 0.000000 \n",
"1502 0.000000 \n",
"1503 0.100000 \n",
"... ... \n",
"16369 0.008333 \n",
"16370 0.018182 \n",
"985 NaN \n",
"1001 NaN \n",
"1002 NaN \n",
"\n",
" Air Temperature Maximum (degF) Air Temperature Minimum (degF) \\\n",
"1499 NaN NaN \n",
"1500 NaN NaN \n",
"1501 NaN NaN \n",
"1502 NaN NaN \n",
"1503 NaN NaN \n",
"... ... ... \n",
"16369 64.615385 35.846154 \n",
"16370 NaN NaN \n",
"985 NaN NaN \n",
"1001 NaN NaN \n",
"1002 NaN NaN \n",
"\n",
" Air Temperature Average (degF) Precipitation Increment (in) \n",
"1499 NaN 0.000000 \n",
"1500 NaN 0.000000 \n",
"1501 NaN 0.000000 \n",
"1502 NaN 0.100000 \n",
"1503 NaN 0.000000 \n",
"... ... ... \n",
"16369 47.923077 0.009091 \n",
"16370 NaN NaN \n",
"985 NaN NaN \n",
"1001 NaN NaN \n",
"1002 NaN NaN \n",
"\n",
"[16371 rows x 12 columns]"
]
},
"metadata": {
"tags": []
},
"execution_count": 17
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "hLwYFKBmuFlk"
},
"source": [
"We can see that the new dataset appended the avalanches columns that we selected into the dataset of weather measurements. Since all occurrences of avalanches are identified with an `Yes` in the column `Avalanche` we can fill the rows with missing data with a `No`, to identify that there was not an avalanche in that date/location. Also we fill missing values with zero in the column of avalanche numbers, for the same reason."
]
},
{
"cell_type": "code",
"metadata": {
"id": "Zsg6LYohZQuF",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "ee978618-a9eb-427b-8f49-206992425921"
},
"source": [
"# Fill null values of avalanche numbers and categorical\n",
"\n",
"FullData['Avalanche'].fillna('No', inplace=True)\n",
"FullData['avi_number'].fillna(0, inplace=True)\n",
"\n",
"# Check information about merged dataset\n",
"\n",
"FullData.info()"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"\n",
"Int64Index: 16371 entries, 1499 to 1002\n",
"Data columns (total 12 columns):\n",
" # Column Non-Null Count Dtype \n",
"--- ------ -------------- ----- \n",
" 0 avi_mark 847 non-null object \n",
" 1 avi_date 1498 non-null datetime64[ns]\n",
" 2 Avalanche 16371 non-null object \n",
" 3 avi_number 16371 non-null float64 \n",
" 4 avi_type 1499 non-null object \n",
" 5 Date 16368 non-null datetime64[ns]\n",
" 6 Snow Water Equivalent (in) Start of Day Values 16368 non-null float64 \n",
" 7 Precipitation Accumulation (in) Start of Day Values 16368 non-null float64 \n",
" 8 Air Temperature Maximum (degF) 15652 non-null float64 \n",
" 9 Air Temperature Minimum (degF) 15649 non-null float64 \n",
" 10 Air Temperature Average (degF) 15652 non-null float64 \n",
" 11 Precipitation Increment (in) 16367 non-null float64 \n",
"dtypes: datetime64[ns](2), float64(7), object(3)\n",
"memory usage: 1.6+ MB\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "iawByAnFukch"
},
"source": [
"In the table above we can see that 1499 avalanches matched the existing data on weather for the locations that were provided.\n",
"\n",
"There are three avalanches entries that didn't have a date assigned for them, so we will delete those entries to avoid issues in analyzing the data."
]
},
{
"cell_type": "code",
"metadata": {
"id": "cD2Up8WjMn2B",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "727b8579-3d3e-4a36-e818-1ec8c82b57ac"
},
"source": [
"FullData.dropna(subset=['Date'], inplace=True)\n",
"\n",
"FullData.info()"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"\n",
"Int64Index: 16368 entries, 1499 to 16370\n",
"Data columns (total 12 columns):\n",
" # Column Non-Null Count Dtype \n",
"--- ------ -------------- ----- \n",
" 0 avi_mark 845 non-null object \n",
" 1 avi_date 1496 non-null datetime64[ns]\n",
" 2 Avalanche 16368 non-null object \n",
" 3 avi_number 16368 non-null float64 \n",
" 4 avi_type 1496 non-null object \n",
" 5 Date 16368 non-null datetime64[ns]\n",
" 6 Snow Water Equivalent (in) Start of Day Values 16368 non-null float64 \n",
" 7 Precipitation Accumulation (in) Start of Day Values 16368 non-null float64 \n",
" 8 Air Temperature Maximum (degF) 15652 non-null float64 \n",
" 9 Air Temperature Minimum (degF) 15649 non-null float64 \n",
" 10 Air Temperature Average (degF) 15652 non-null float64 \n",
" 11 Precipitation Increment (in) 16367 non-null float64 \n",
"dtypes: datetime64[ns](2), float64(7), object(3)\n",
"memory usage: 1.6+ MB\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "6TQe7zbQdFd_"
},
"source": [
"Since the dataset is very unbalanced we can remove all data before 2014, since we don't have information about avalanches prior to that.\n",
"\n"
]
},
{
"cell_type": "code",
"metadata": {
"id": "rWHAs8fYfPb2",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "6c0cab7d-365a-4cfa-a49a-cf10d4c9a94b"
},
"source": [
"FullData = FullData[FullData['Date'].dt.year >= 2014]\n",
"\n",
"FullData.info()"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"\n",
"Int64Index: 3492 entries, 14375 to 16370\n",
"Data columns (total 12 columns):\n",
" # Column Non-Null Count Dtype \n",
"--- ------ -------------- ----- \n",
" 0 avi_mark 845 non-null object \n",
" 1 avi_date 1496 non-null datetime64[ns]\n",
" 2 Avalanche 3492 non-null object \n",
" 3 avi_number 3492 non-null float64 \n",
" 4 avi_type 1496 non-null object \n",
" 5 Date 3492 non-null datetime64[ns]\n",
" 6 Snow Water Equivalent (in) Start of Day Values 3492 non-null float64 \n",
" 7 Precipitation Accumulation (in) Start of Day Values 3492 non-null float64 \n",
" 8 Air Temperature Maximum (degF) 3491 non-null float64 \n",
" 9 Air Temperature Minimum (degF) 3491 non-null float64 \n",
" 10 Air Temperature Average (degF) 3491 non-null float64 \n",
" 11 Precipitation Increment (in) 3491 non-null float64 \n",
"dtypes: datetime64[ns](2), float64(7), object(3)\n",
"memory usage: 354.7+ KB\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "3TGRt_WmvTUU"
},
"source": [
"With that the pre-processing section is finished and there is a unique dataframe which we can analyze to check if the existing weather data shows correlation with avalanche occurrences."
]
},
{
"cell_type": "code",
"metadata": {
"id": "RZYlZgEuffju",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 377
},
"outputId": "d8c0ae1e-c34d-4283-9ada-8a65225608ff"
},
"source": [
"FullData.head()"
],
"execution_count": null,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/html": [
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
avi_mark
\n",
"
avi_date
\n",
"
Avalanche
\n",
"
avi_number
\n",
"
avi_type
\n",
"
Date
\n",
"
Snow Water Equivalent (in) Start of Day Values
\n",
"
Precipitation Accumulation (in) Start of Day Values
\n",
"
Air Temperature Maximum (degF)
\n",
"
Air Temperature Minimum (degF)
\n",
"
Air Temperature Average (degF)
\n",
"
Precipitation Increment (in)
\n",
"
\n",
" \n",
" \n",
"
\n",
"
14375
\n",
"
NaN
\n",
"
NaT
\n",
"
No
\n",
"
0.0
\n",
"
NaN
\n",
"
2014-01-01
\n",
"
8.923077
\n",
"
9.615385
\n",
"
27.384615
\n",
"
11.000000
\n",
"
19.461538
\n",
"
0.207692
\n",
"
\n",
"
\n",
"
14376
\n",
"
NaN
\n",
"
NaT
\n",
"
No
\n",
"
0.0
\n",
"
NaN
\n",
"
2014-01-02
\n",
"
9.146154
\n",
"
9.823077
\n",
"
36.076923
\n",
"
11.230769
\n",
"
21.615385
\n",
"
0.046154
\n",
"
\n",
"
\n",
"
14377
\n",
"
NaN
\n",
"
NaT
\n",
"
No
\n",
"
0.0
\n",
"
NaN
\n",
"
2014-01-03
\n",
"
9.176923
\n",
"
9.869231
\n",
"
37.307692
\n",
"
16.076923
\n",
"
25.692308
\n",
"
0.023077
\n",
"
\n",
"
\n",
"
14378
\n",
"
NaN
\n",
"
NaT
\n",
"
No
\n",
"
0.0
\n",
"
NaN
\n",
"
2014-01-04
\n",
"
9.238462
\n",
"
9.892308
\n",
"
26.153846
\n",
"
-2.230769
\n",
"
12.307692
\n",
"
0.161538
\n",
"
\n",
"
\n",
"
14379
\n",
"
NaN
\n",
"
NaT
\n",
"
No
\n",
"
0.0
\n",
"
NaN
\n",
"
2014-01-05
\n",
"
9.500000
\n",
"
10.053846
\n",
"
7.153846
\n",
"
-5.076923
\n",
"
0.384615
\n",
"
0.069231
\n",
"
\n",
" \n",
"
\n",
"
"
],
"text/plain": [
" avi_mark avi_date Avalanche avi_number avi_type Date \\\n",
"14375 NaN NaT No 0.0 NaN 2014-01-01 \n",
"14376 NaN NaT No 0.0 NaN 2014-01-02 \n",
"14377 NaN NaT No 0.0 NaN 2014-01-03 \n",
"14378 NaN NaT No 0.0 NaN 2014-01-04 \n",
"14379 NaN NaT No 0.0 NaN 2014-01-05 \n",
"\n",
" Snow Water Equivalent (in) Start of Day Values \\\n",
"14375 8.923077 \n",
"14376 9.146154 \n",
"14377 9.176923 \n",
"14378 9.238462 \n",
"14379 9.500000 \n",
"\n",
" Precipitation Accumulation (in) Start of Day Values \\\n",
"14375 9.615385 \n",
"14376 9.823077 \n",
"14377 9.869231 \n",
"14378 9.892308 \n",
"14379 10.053846 \n",
"\n",
" Air Temperature Maximum (degF) Air Temperature Minimum (degF) \\\n",
"14375 27.384615 11.000000 \n",
"14376 36.076923 11.230769 \n",
"14377 37.307692 16.076923 \n",
"14378 26.153846 -2.230769 \n",
"14379 7.153846 -5.076923 \n",
"\n",
" Air Temperature Average (degF) Precipitation Increment (in) \n",
"14375 19.461538 0.207692 \n",
"14376 21.615385 0.046154 \n",
"14377 25.692308 0.023077 \n",
"14378 12.307692 0.161538 \n",
"14379 0.384615 0.069231 "
]
},
"metadata": {
"tags": []
},
"execution_count": 21
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "_60H9-nT2bxZ"
},
"source": [
"## Methods for Analyzing Data"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "htwpYkVr3DHd"
},
"source": [
"### Practical Analysis"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "bVVopWUN3J_X"
},
"source": [
"For Practical Analysis, we look into Time Series Plots, Distribution Plots, and Distribution Plots by Avalanche Type to identify a relationship between avalanche occurrences and the weather. We look at descriptive statistics information about the data, compared time with different weather variables, and weather variables with the avalanche count. Then we did something similar, but with types of avalanches."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "FgkGe1kU3RjH"
},
"source": [
"### Linear Regression"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "xpvKS9E-3XHB"
},
"source": [
"Linear regression was something that we attempted to use in order to determine if there was a direct correlation between each weather variable and the occurence of avalanches. Linear regression requires a set of independent variables, x, and a dependent variable, y, to determine if there is a linear relationship between them. If there is a linear relationship, it can be used to predict the future occurence of the dependent variable. In this scenario, the variable 'avi_number' is the dependent variable that we are attempting to predict, and the weather variables, 'Snow Water Equivalent (in) Start of Day Values', 'Precipitation Accumulation (in) Start of Day Values',\t'Air Temperature Average (degF)',\t'Air Temperature Maximum (degF)',\t'Air Temperature Minimum (degF)',\tand 'Precipitation Increment' were each used as the various independent variables. \n",
"\n",
"We used the [NumPy](https://numpy.org) and [skicit-learn](https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LinearRegression.html) packages in order to implement linear regression. We manipulated the 'avi_number' data in multiple ways to search for an R2 score that would be significant enough for us to pursue linear regression, and based off our findings, decided it was not an appropriate method for calculating correlation. \n"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "TQ_7vtHyJ7il"
},
"source": [
"### Logistic Regression"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "Hp9QxcQCKQFO"
},
"source": [
"Logistic regression was a model that we wanted to look into further because it utilizes avalanche count as a binary variable (0 for no avalanche/1 for an avalanche occurence). This is ideal because we are trying to predict the occurence of an avalanche under different weather variables. \n",
"\n",
"We used the [scikit-learn](https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html) package to implement logistic regression. The variable 'AvCount' was created as our binary dependent variable because we are trying to predict it. The highest accuracy model was generated when we used all of the available weather variables, 'Snow Water Equivalent (in) Start of Day Values', 'Precipitation Accumulation (in) Start of Day Values',\t'Air Temperature Average (degF)',\t'Air Temperature Maximum (degF)',\t'Air Temperature Minimum (degF)',\tand 'Precipitation Increment'. More information about the logistic regression model that we implemented is available [here](https://www.datacamp.com/community/tutorials/understanding-logistic-regression-python).\n",
"\n",
"This method of logistic regression differs from the logistic regression model in the machine learning portion of our results. That model is generated using PyCaret which a machine learning library and doesn't require as much code to run and concludes the most important predictors without intervention from the coder."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "_TnoMwNF27DO"
},
"source": [
"### Machine Learning"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "Up35Eaxz2Xw9"
},
"source": [
"We used PyCaret which is an open-source, low-code machine learning library in Python. It allows us to check the accuracy of multiple methods at ones. The methods included are both categorical and regression. From there we took the columns Avalanche, Snow Water Equivalent, Precipitation Accumilation, Air Temperature Maximum, Ait Temperature Minimum, Air Temperature Average, and Precipitation Increment. Then we set up our target to be Avalanche where No became 0 and Yes became 1. It then shows you defaults like the [fold number](https://www.openml.org/a/estimation-procedus/7), which is the original sample is randomly partitioned into k equal size subsamples for training and testing. Its default for testing and training is 70:30. Once that step is done, you compare models and it shows you the accuracy from highest to lowest. We then took three models from different accuracies, but still high, they were Cat Boost, extreme Gradient Boost, and Logistic Regression. We then tuned each model, evaluated it using a [Confusion Matrix](https://pycaret.org/plot-model/), and included a portion where it plots a Feature Importance Plot and a heat map."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "DUy3OgBc0iEX"
},
"source": [
"# Results and Discussion"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "XlEmNMGMCq4R"
},
"source": [
"## Practical Analysis\n",
"\n",
"In the practical analysis we will use statistical and graphical techniques to see if we can identify a relationship between avalanche occurrences and the weather.\n",
"\n",
"First we start checking descriptive statistics information about the data, to see if there is anything strange and get a sense of the distribution of the data."
]
},
{
"cell_type": "code",
"metadata": {
"id": "NkcKvVbMCgwI",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 266
},
"outputId": "d2a54573-c1c9-4111-bf01-c3c5e650bf73"
},
"source": [
"# Descriptive Statistics of numerical variables\n",
"\n",
"FullData.describe().transpose()"
],
"execution_count": null,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/html": [
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
count
\n",
"
mean
\n",
"
std
\n",
"
min
\n",
"
25%
\n",
"
50%
\n",
"
75%
\n",
"
max
\n",
"
\n",
" \n",
" \n",
"
\n",
"
avi_number
\n",
"
3492.0
\n",
"
0.995132
\n",
"
2.836968
\n",
"
0.000000
\n",
"
0.000000
\n",
"
0.000000
\n",
"
1.000000
\n",
"
100.000000
\n",
"
\n",
"
\n",
"
Snow Water Equivalent (in) Start of Day Values
\n",
"
3492.0
\n",
"
10.647652
\n",
"
8.890274
\n",
"
0.000000
\n",
"
0.223077
\n",
"
11.042308
\n",
"
18.063462
\n",
"
28.484615
\n",
"
\n",
"
\n",
"
Precipitation Accumulation (in) Start of Day Values
\n",
"
3492.0
\n",
"
19.511723
\n",
"
9.630722
\n",
"
0.000000
\n",
"
13.292308
\n",
"
19.676923
\n",
"
26.450000
\n",
"
39.138462
\n",
"
\n",
"
\n",
"
Air Temperature Maximum (degF)
\n",
"
3491.0
\n",
"
48.189433
\n",
"
12.888759
\n",
"
7.153846
\n",
"
39.615385
\n",
"
48.230769
\n",
"
57.769231
\n",
"
75.153846
\n",
"
\n",
"
\n",
"
Air Temperature Minimum (degF)
\n",
"
3491.0
\n",
"
26.276734
\n",
"
12.210477
\n",
"
-13.846154
\n",
"
18.307692
\n",
"
26.916667
\n",
"
35.538462
\n",
"
48.923077
\n",
"
\n",
"
\n",
"
Air Temperature Average (degF)
\n",
"
3491.0
\n",
"
36.078471
\n",
"
12.131301
\n",
"
-1.846154
\n",
"
27.884615
\n",
"
36.230769
\n",
"
45.615385
\n",
"
60.230769
\n",
"
\n",
"
\n",
"
Precipitation Increment (in)
\n",
"
3491.0
\n",
"
0.092666
\n",
"
0.148213
\n",
"
0.000000
\n",
"
0.007692
\n",
"
0.038462
\n",
"
0.107692
\n",
"
1.523077
\n",
"
\n",
" \n",
"
\n",
"
"
],
"text/plain": [
" count mean \\\n",
"avi_number 3492.0 0.995132 \n",
"Snow Water Equivalent (in) Start of Day Values 3492.0 10.647652 \n",
"Precipitation Accumulation (in) Start of Day Va... 3492.0 19.511723 \n",
"Air Temperature Maximum (degF) 3491.0 48.189433 \n",
"Air Temperature Minimum (degF) 3491.0 26.276734 \n",
"Air Temperature Average (degF) 3491.0 36.078471 \n",
"Precipitation Increment (in) 3491.0 0.092666 \n",
"\n",
" std min \\\n",
"avi_number 2.836968 0.000000 \n",
"Snow Water Equivalent (in) Start of Day Values 8.890274 0.000000 \n",
"Precipitation Accumulation (in) Start of Day Va... 9.630722 0.000000 \n",
"Air Temperature Maximum (degF) 12.888759 7.153846 \n",
"Air Temperature Minimum (degF) 12.210477 -13.846154 \n",
"Air Temperature Average (degF) 12.131301 -1.846154 \n",
"Precipitation Increment (in) 0.148213 0.000000 \n",
"\n",
" 25% 50% \\\n",
"avi_number 0.000000 0.000000 \n",
"Snow Water Equivalent (in) Start of Day Values 0.223077 11.042308 \n",
"Precipitation Accumulation (in) Start of Day Va... 13.292308 19.676923 \n",
"Air Temperature Maximum (degF) 39.615385 48.230769 \n",
"Air Temperature Minimum (degF) 18.307692 26.916667 \n",
"Air Temperature Average (degF) 27.884615 36.230769 \n",
"Precipitation Increment (in) 0.007692 0.038462 \n",
"\n",
" 75% max \n",
"avi_number 1.000000 100.000000 \n",
"Snow Water Equivalent (in) Start of Day Values 18.063462 28.484615 \n",
"Precipitation Accumulation (in) Start of Day Va... 26.450000 39.138462 \n",
"Air Temperature Maximum (degF) 57.769231 75.153846 \n",
"Air Temperature Minimum (degF) 35.538462 48.923077 \n",
"Air Temperature Average (degF) 45.615385 60.230769 \n",
"Precipitation Increment (in) 0.107692 1.523077 "
]
},
"metadata": {
"tags": []
},
"execution_count": 22
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "91UBVOGPv385"
},
"source": [
"Despite the wide temperature distribution existing in the data, there are no signs of anomalies in the measurement that should be removed from the dataset. Let's also check some information about the categorical variables below."
]
},
{
"cell_type": "code",
"metadata": {
"id": "jOHObO95C0km",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 142
},
"outputId": "a033d4ed-f21c-459a-e8ac-b5b17b5eddd2"
},
"source": [
"# Descriptive Statistics of categorical variables\n",
"\n",
"FullData.describe(include='O').transpose()"
],
"execution_count": null,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/html": [
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
count
\n",
"
unique
\n",
"
top
\n",
"
freq
\n",
"
\n",
" \n",
" \n",
"
\n",
"
avi_mark
\n",
"
845
\n",
"
94
\n",
"
-1
\n",
"
124
\n",
"
\n",
"
\n",
"
Avalanche
\n",
"
3492
\n",
"
2
\n",
"
No
\n",
"
1996
\n",
"
\n",
"
\n",
"
avi_type
\n",
"
1496
\n",
"
2
\n",
"
WL
\n",
"
1009
\n",
"
\n",
" \n",
"
\n",
"
"
],
"text/plain": [
" count unique top freq\n",
"avi_mark 845 94 -1 124\n",
"Avalanche 3492 2 No 1996\n",
"avi_type 1496 2 WL 1009"
]
},
"metadata": {
"tags": []
},
"execution_count": 23
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "hIG0WGkNwPsE"
},
"source": [
"We can see that the Wet Loose avalanches represent more than two thirds of all avalanches between 2014 and today. Also the dataset seems well balanced between occurences of avalanches.\n",
"\n",
"Let's check some time series plots about the weather and avalanches."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "IpOpzDYcFFol"
},
"source": [
"### Time Series Plots"
]
},
{
"cell_type": "code",
"metadata": {
"id": "ep2bDTEmnABr",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 323
},
"outputId": "6bf9c4f3-8f59-4197-f133-157fa37952bf"
},
"source": [
"# Group data by date to improve plot quality\n",
"\n",
"GroupedData = FullData.groupby('Date').agg({'Snow Water Equivalent (in) Start of Day Values':'mean',\n",
" 'Precipitation Accumulation (in) Start of Day Values':'mean',\n",
" 'Air Temperature Average (degF)':'mean', \n",
" 'Air Temperature Maximum (degF)':'mean', \n",
" 'Air Temperature Minimum (degF)':'mean',\n",
" 'Precipitation Increment (in)':'mean',\n",
" 'avi_number':'sum'}).reset_index()\n",
"\n",
"GroupedData.head()"
],
"execution_count": null,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/html": [
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
Date
\n",
"
Snow Water Equivalent (in) Start of Day Values
\n",
"
Precipitation Accumulation (in) Start of Day Values
\n",
"\n",
""
]
},
"metadata": {
"tags": []
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "BB4ynOxtiIBq"
},
"source": [
"By looking at the time series data we can identify a pattern. It seems that most avalanches happen when temperatures are rising. This makes a lot of sense because it is when there is most snow in the mountains, and with the increase in temperature this snow starts to melt. Once it is all melted there is not enough snow to generate an avalanche.\n",
"\n",
"By having those initial insights we can look more in details to each one of the weather variables measured, and see if it is correlated with avalanche occurrences. To do that we can create a distribution plot and split the data from the days when there was an avalanche and when there wasn't, to compare the distributions."
]
},
{
"cell_type": "code",
"metadata": {
"id": "g9hyQ6k-xL4Z",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 542
},
"outputId": "06b92ff1-c693-420b-a0f6-3c21f6030223"
},
"source": [
"# Create figure with secondary y-axis\n",
"fig = make_subplots(specs=[[{\"secondary_y\": True}]])\n",
"\n",
"# Add traces\n",
"fig.add_trace(\n",
" go.Scatter(x=GroupedData['Date'], y=GroupedData['Precipitation Accumulation (in) Start of Day Values'], name=\"Precipitation Accumulation\"),\n",
" secondary_y=False, \n",
")\n",
"\n",
"fig.add_trace(\n",
" go.Scatter(x=GroupedData['Date'], y=GroupedData['avi_number'],mode= 'markers', name=\"Number of Avalanches\"),\n",
" secondary_y=True,\n",
")\n",
"\n",
"# Add figure title\n",
"fig.update_layout(\n",
" title_text=\"2-Axis Plot (Precipitation Accumulation)\"\n",
")\n",
"\n",
"# Set x-axis title\n",
"fig.update_xaxes(title_text=\"Date\")\n",
"\n",
"# Set y-axes titles\n",
"fig.update_yaxes(title_text=\"Precipitation Accumulation\", secondary_y=False)\n",
"fig.update_yaxes(title_text=\"Number of Avalanches\", secondary_y=True)\n",
"\n",
"fig.show()"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"text/html": [
"\n",
"\n",
"\n",
"
\n",
"\n",
""
]
},
"metadata": {
"tags": []
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "Aq9q3YcSf_Xq"
},
"source": [
"We can clearly see that there is a significant difference between the results of days when there was an avalanche and of then there was none and that it should be detected by a prediction model.\n",
"\n",
"Now adding some critical thinking on those results, it is important to remember that the data that is being used is an average of all locations, and not the measurement in the specific location where the avalanche happened. So it is better to read the results as the relations between weather data and any incidence of avalanche.\n",
"\n",
"Also we saw that spring months are definitely the ones with most avalanche (probably because the snow is melting). If there is a relation between the season and the snow water equivalent results, it is possible that the results that we are seeing are not directly correlated, but are correlated by a third variable. But in this specific case there is a natural theorical explanation that when there is more snow, there is a higher instability in the mountains that can lead to avalanches.\n",
"\n",
"More details on Snow Water Equivalent:\n",
"[Link](https://www.nrcs.usda.gov/wps/portal/nrcs/detail/null/?cid=nrcseprd1314833)"
]
},
{
"cell_type": "code",
"metadata": {
"id": "radcs6OkDTW5",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 542
},
"outputId": "77d125ff-6deb-4223-b457-e7a39e1309e0"
},
"source": [
"# Create histogram / distribution visualization for the different variables\n",
"\n",
"px.histogram(FullData,\n",
" x='Precipitation Accumulation (in) Start of Day Values',\n",
" color='Avalanche',\n",
" marginal='box')"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"text/html": [
"\n",
"\n",
"\n",
"
\n",
"\n",
""
]
},
"metadata": {
"tags": []
}
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "88-PF426iF4Q",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 542
},
"outputId": "fab497e6-adce-4acd-9768-797c431b4c6b"
},
"source": [
"# Create histogram / distribution visualization for the different variables\n",
"\n",
"variable = 'Air Temperature Average (degF)' #@param ['Location', 'Date', 'Snow Water Equivalent (in) Start of Day Values','Precipitation Accumulation (in) Start of Day Values','Air Temperature Maximum (degF)', 'Air Temperature Minimum (degF)','Air Temperature Average (degF)', 'Precipitation Increment (in)','avi_mark', 'avi_date', 'Avalanche', 'avi_number']\n",
"\n",
"fig = px.histogram(AvalanchesOnly,\n",
" x=variable,\n",
" color='avi_type',\n",
" marginal='box')\n",
"\n",
"fig.show()"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"text/html": [
"\n",
"\n",
"\n",
"
\n",
" \n",
" \n",
" \n",
" \n",
" \n",
"
\n",
"\n",
""
]
},
"metadata": {
"tags": []
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "WN461FtJz_EY"
},
"source": [
"Here there weren't many big differences in the measured variables' distribution. The one specific observation is that it seems that Wet Slab avalanches has a slightly higher median for the Minimum, Maximum and Average temperatures than the Wet Loose ones. It is possible to test that difference in the medians or means to understand if that is statistically significant.\n",
"\n"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "fHmyv-a_IaDb"
},
"source": [
"## Linear Regression"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "LwJ2u9-tMufH"
},
"source": [
"The following code shows our implementation of linear regression and why it did not work. The 'avi_number' data was our biggest obstacle because of how sparse the data was. Multiple attempts were made in order to account for this sparseness; however, none of them generated on R^2 score that we felt was satisfactory enough to continue working on linear regression."
]
},
{
"cell_type": "code",
"metadata": {
"id": "qQw4-UBtPaF1"
},
"source": [
"#Creating a copy of the data without null values\n",
"NoNullGroupedData = GroupedData.dropna()"
],
"execution_count": null,
"outputs": []
},
{
"cell_type": "markdown",
"metadata": {
"id": "IRpndxuw8OZY"
},
"source": [
"This is the first attempt to test for a correlation between the weather variable, Snow Water Equivalent, and avalanche count. This attempt uses the data as is, with no adjustments made to avalanche count. "
]
},
{
"cell_type": "code",
"metadata": {
"id": "LCopvMBTMC3P",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "04c80476-90a9-4932-c654-57fc19501db9"
},
"source": [
"#creating our x and y variables \n",
"x1 = NoNullGroupedData.iloc[:,[3]].values.reshape(-1, 1) #Snow Water Equivalent\n",
"y1 = NoNullGroupedData.iloc[:,[9]].values.reshape(-1, 1) #avi_number\n",
"\n",
"# Model initialization\n",
"regression_model_one = LinearRegression()\n",
"# Fit the data(train the model)\n",
"regression_model_one.fit(x1, y1)\n",
"#Calculate and print R2 score\n",
"r_sq_one = regression_model_one.score(x1,y1)\n",
"print('R2 score: ', r_sq_one)"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"R2 score: 0.07665051806996348\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "FTS9kudU9hBT"
},
"source": [
"Ideally, the R2 score should be as close to 1 as possible. The low R2 score shows that the data, in this form, does not capture the relationship between Snow Water Equivalence and the occurence of avalanches\n"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "poAdVJ0Y9t1y"
},
"source": [
"This is the second attempt to test for a correlation between the weather variable, Snow Water Equivalent, and avalanche count. In this attempt, we made avalanche count a binary variable (0 for no/1 for yes).\n",
"\n"
]
},
{
"cell_type": "code",
"metadata": {
"id": "B70zHsTz99B6"
},
"source": [
"#making new column for avalanche data as a binary count\n",
"NoNullGroupedData.loc[NoNullGroupedData['avi_number'] == 0, 'AvCount'] = 0 \n",
"NoNullGroupedData.loc[NoNullGroupedData['avi_number'] >= 1, 'AvCount'] = 1"
],
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"id": "qVxAiSEP-EL4",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "dafeaa28-31f9-47b5-cee7-27691a529a98"
},
"source": [
"#creating our x and y variables \n",
"x2 = NoNullGroupedData.iloc[:,[3]].values.reshape(-1, 1) #Snow Water Equivalent\n",
"y2 = NoNullGroupedData.iloc[:,[10]].values.reshape(-1, 1) #AvCount\n",
"\n",
"# Model initialization\n",
"regression_model_two = LinearRegression()\n",
"# Fit the data(train the model)\n",
"regression_model_two.fit(x2, y2)\n",
"#Calculate and print R2 score\n",
"r_sq_two = regression_model_two.score(x2,y2)\n",
"print('R2 score: ', r_sq_two)"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"R2 score: 0.2755908525407361\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "KDMVno-t-8kG"
},
"source": [
"The R2 score generated is helpful in understanding there is a relationship between 'Snow Water Equivalent' and the occurence of an avalalanche. However, it is not significant enough for us to pursue it further. "
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "ptrRGkS4HRtD"
},
"source": [
"This is the third attempt to test for a correlation between the weather variable, Snow Water Equivalent, and avalanche count. In this attempt, we removed all of the months after May because the majority of avalanches occur at this time. We also could have included the months November and December; however, the smaller sample worked just as well."
]
},
{
"cell_type": "code",
"metadata": {
"id": "y9vLT4FqAc7a"
},
"source": [
"#Only including data between January and May\n",
"NoNullGroupedData = NoNullGroupedData[NoNullGroupedData['Date'].dt.month <= 5]"
],
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"id": "gspRoMiLAeaX",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "db5c7569-9491-45de-c48c-811ceca37c9a"
},
"source": [
"#creating our x and y variables \n",
"x3 = NoNullGroupedData.iloc[:,[3]].values.reshape(-1, 1) #Snow Water Equivalent\n",
"y3 = NoNullGroupedData.iloc[:,[9]].values.reshape(-1, 1) #avi_number\n",
"\n",
"# Model initialization\n",
"regression_model_three = LinearRegression()\n",
"# Fit the data(train the model)\n",
"regression_model_three.fit(x3, y3)\n",
"#Calculate and print R2 score\n",
"r_sq_three = regression_model_three.score(x3,y3)\n",
"print('R2 score: ', r_sq_three)"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"R2 score: 0.020505479899889267\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "ipg2zqGsAtbp"
},
"source": [
"The low R2 score shows that the data, in this form, does not capture the relationship between Snow Water Equivalence and the occurence of avalanches"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "dpR53Y7UFXpv"
},
"source": [
"## Logistic Regression"
]
},
{
"cell_type": "code",
"metadata": {
"id": "tJjQRl7TZC-m",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 296
},
"outputId": "9ba78ab4-ae39-44f8-bbfc-38d9659e2cc8"
},
"source": [
"#split dataset in features and target variable\n",
"X = NoNullGroupedData[['Day of Year',\n",
" 'Snow Water Equivalent (in) Start of Day Values', \n",
" 'Precipitation Accumulation (in) Start of Day Values',\n",
" 'Precipitation Increment (in)',\n",
" 'Air Temperature Average (degF)',\n",
" 'Air Temperature Maximum (degF)',\n",
" 'Air Temperature Minimum (degF)']]\n",
"y = NoNullGroupedData['AvCount']\n",
"\n",
"#splitting data into a testing set and training set\n",
"X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.25,random_state=0)\n",
"\n",
"#creating an instance of the logistic regression model and fitting it with data\n",
"logistic_regression= LogisticRegression()\n",
"logistic_regression.fit(X_train,y_train)\n",
"y_pred=logistic_regression.predict(X_test)\n",
"\n",
"#confusion matrix used to visualize the predictions\n",
"confusion_matrix = pd.crosstab(y_test, y_pred, rownames=['Actual'], colnames=['Predicted'])\n",
"sn.heatmap(confusion_matrix, annot=True, fmt='g')\n",
"\n",
"#displaying the confusion matrix and accuracy of the model\n",
"print('Accuracy: ',metrics.accuracy_score(y_test, y_pred))\n",
"plt.show()"
],
"execution_count": null,
"outputs": [
{
"output_type": "stream",
"text": [
"Accuracy: 0.7283018867924528\n"
],
"name": "stdout"
},
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
"
"
]
},
"metadata": {
"tags": [],
"needs_background": "light"
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "Buf4iNsL6ugH"
},
"source": [
"In this confusion matrix, the 119 and 74 are actual predictions, and 40 and 32 are incorrect predictions. The model has as accuracy of approximately .73."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "KEU1OfPB0-xG"
},
"source": [
"## Prediction Model"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "zuooMMfYX5So"
},
"source": [
"We were unable to get this code to run."
]
},
{
"cell_type": "code",
"metadata": {
"id": "muoTRjx7S8OT",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 292
},
"outputId": "8063f542-30ad-4044-9c81-daa14866bb16"
},
"source": [
"ModelDataset = FullData.copy()\n",
"ModelDataset2 = FullData.copy()\n",
"d=ModelDataset['Date']\n",
"ModelDataset.insert(1,'Year',d.dt.year)\n",
"ModelDataset.insert(2,'Day of Year',d.dt.dayofyear)\n",
"ModelDataset.drop(['Date','avi_mark','avi_date', 'avi_number', 'avi_type'], axis=1, inplace=True)\n",
"ModelDataset2.drop(['avi_mark','avi_date', 'avi_type'], axis=1, inplace=True)\n",
"ModelDataset2.insert(1,'Year',d.dt.year)\n",
"ModelDataset2.insert(2,'Day of Year',d.dt.dayofyear)\n",
"del d\n",
"\n",
"ModelDataset.head()"
],
"execution_count": null,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/html": [
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
Year
\n",
"
Day of Year
\n",
"
Avalanche
\n",
"
Snow Water Equivalent (in) Start of Day Values
\n",
"
Precipitation Accumulation (in) Start of Day Values
\n",
"
Air Temperature Maximum (degF)
\n",
"
Air Temperature Minimum (degF)
\n",
"
Air Temperature Average (degF)
\n",
"
Precipitation Increment (in)
\n",
"
\n",
" \n",
" \n",
"
\n",
"
14375
\n",
"
2014
\n",
"
1
\n",
"
No
\n",
"
8.923077
\n",
"
9.615385
\n",
"
27.384615
\n",
"
11.000000
\n",
"
19.461538
\n",
"
0.207692
\n",
"
\n",
"
\n",
"
14376
\n",
"
2014
\n",
"
2
\n",
"
No
\n",
"
9.146154
\n",
"
9.823077
\n",
"
36.076923
\n",
"
11.230769
\n",
"
21.615385
\n",
"
0.046154
\n",
"
\n",
"
\n",
"
14377
\n",
"
2014
\n",
"
3
\n",
"
No
\n",
"
9.176923
\n",
"
9.869231
\n",
"
37.307692
\n",
"
16.076923
\n",
"
25.692308
\n",
"
0.023077
\n",
"
\n",
"
\n",
"
14378
\n",
"
2014
\n",
"
4
\n",
"
No
\n",
"
9.238462
\n",
"
9.892308
\n",
"
26.153846
\n",
"
-2.230769
\n",
"
12.307692
\n",
"
0.161538
\n",
"
\n",
"
\n",
"
14379
\n",
"
2014
\n",
"
5
\n",
"
No
\n",
"
9.500000
\n",
"
10.053846
\n",
"
7.153846
\n",
"
-5.076923
\n",
"
0.384615
\n",
"
0.069231
\n",
"
\n",
" \n",
"
\n",
"
"
],
"text/plain": [
" Year ... Precipitation Increment (in)\n",
"14375 2014 ... 0.207692\n",
"14376 2014 ... 0.046154\n",
"14377 2014 ... 0.023077\n",
"14378 2014 ... 0.161538\n",
"14379 2014 ... 0.069231\n",
"\n",
"[5 rows x 9 columns]"
]
},
"metadata": {
"tags": []
},
"execution_count": 76
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "vlhsXyR3TXvE"
},
"source": [
"Once our dataset is ready, we can define what our target variable below. For the Confusion Matrix Charts 0 is equal to no avalanches and 1 is equal to yes avalanches.\n"
]
},
{
"cell_type": "code",
"metadata": {
"id": "LTx3VNDmTdQJ",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 1000,
"referenced_widgets": [
"499d8582265046e2970e453489ba6031",
"fbf81571f4cb4c059a3e5a66b7ef2221",
"4347d82e97784a83a087906c398d0e1f",
"a6c8551cb31645318ee511bd4a1e3859",
"7630cddea594473e9538bef58294d8c2",
"169d6da9acf647d6bd4f509a5fdd50ba"
]
},
"outputId": "f5fe5c38-5a02-48c5-af6b-7fba81584db2"
},
"source": [
"Model = setup(ModelDataset, target='Avalanche', numeric_features=['Year'])\n"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"text/html": [
"
"
]
},
"metadata": {
"tags": []
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "zh6wrvNxZFF0"
},
"source": [
"The confusion matrix shows us all the data points that were tested to verify the quality of the model. From the tested data points:\n",
"\n",
"\n",
"* 545 points were correctly classified as days without avalanches\n",
"* 47 days had an avalanche predicted but none happened\n",
"* In 405 days avalanches were predicted and they really happened\n",
"* In 51 days there were unpredicted avalanches\n",
"\n",
"\n",
"\n"
]
},
{
"cell_type": "code",
"metadata": {
"id": "6G086_FrXTJ0",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 418
},
"outputId": "1f7dbbc8-d600-4eae-9ffe-2687e463da76"
},
"source": [
"plot_model(XGB, 'feature')"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
"
"
]
},
"metadata": {
"tags": []
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "mR96DWU_FEQB"
},
"source": [
"###Testing and Training"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "4g_e_cfcG5G5"
},
"source": [
"In this section, the data was tested and trained with a 60:40 and 80:20. Previously the data was tested and trained with a 70:30. The reason this is done is to see if the accuracy of the models changes. We look into the same three models as earlier. Included is a link showing all charts together so that it is easier to look through them. It shows all of the same models with similiar accuracies even if the testing and training ratio changes. The three models are Logistic Regression, CatBoost, and Extreme Gradient Boost. All models and testing and training has showed that Snow Water Equivalent is important. https://drive.google.com/file/d/1njKn81QJfTAFwHFyHKL_ljy8a6pCZoV4/view?usp=sharing"
]
},
{
"cell_type": "code",
"metadata": {
"id": "b-sK6pFrFqGJ",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 616,
"referenced_widgets": [
"afd584fbc8a24f1baaf7ffd336839df5",
"9e412e3be60c4a65a00a62cc0900524c",
"650fd903437f42138427259ae89385c0",
"adc3728c289f4410b594a4212f5e729c",
"dd30b884e50b47caa2c7eb30d7d5b397",
"eb3d75a11c1c4575a621c248ccfe0f52",
"60bb4204212e4b65b4e7fdf846790952",
"7ea62fdec17144158b7fb4cfe3c83b21",
"cf5589d70daf40ed94c7ffd89dac990e"
]
},
"outputId": "f6ff91f4-f2a0-4954-baf2-98251e690e23"
},
"source": [
"Model = setup(ModelDataset, target='Avalanche',train_size = 0.6,numeric_features=['Year'])\n",
"compare_models()"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"text/html": [
"
"
],
"text/plain": [
" Accuracy AUC Recall Prec. F1 Kappa MCC\n",
"0 0.9143 0.9732 0.9167 0.8871 0.9016 0.8257 0.8261\n",
"1 0.9143 0.9698 0.8917 0.9068 0.8992 0.8246 0.8247\n",
"2 0.9393 0.9786 0.9333 0.9256 0.9295 0.8762 0.8762\n",
"3 0.9140 0.9696 0.9244 0.8800 0.9016 0.8253 0.8261\n",
"4 0.8817 0.9532 0.8908 0.8413 0.8653 0.7600 0.7610\n",
"5 0.9211 0.9745 0.9160 0.9008 0.9083 0.8392 0.8392\n",
"6 0.9032 0.9679 0.8992 0.8770 0.8880 0.8028 0.8030\n",
"7 0.9068 0.9693 0.9076 0.8780 0.8926 0.8103 0.8107\n",
"8 0.9211 0.9728 0.9412 0.8819 0.9106 0.8402 0.8416\n",
"9 0.9462 0.9905 0.9748 0.9062 0.9393 0.8912 0.8931\n",
"Mean 0.9162 0.9719 0.9195 0.8885 0.9036 0.8295 0.8302\n",
"SD 0.0172 0.0089 0.0243 0.0218 0.0197 0.0349 0.0350"
]
},
"metadata": {
"tags": []
}
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "_5346QYnpG3l",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 1000,
"referenced_widgets": [
"f298fe3f4539460f892778917dca8e35",
"3ae88fa2ea8642d88a4b44091352d1ae",
"268e0463c4b44a1a9811775756d77c41",
"d434264ed45c4de8af5e31ac2901b8c4",
"497321ff3b6849b19b7ab938133316ad",
"cc0fbcf2f77846f4bd1266103d07a1f7",
"b995a9f6bf614675ac2153fc65daa3de"
]
},
"outputId": "6d0aa99a-122a-4774-84f0-032d101b8171"
},
"source": [
"evaluate_model(catboost)"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "f298fe3f4539460f892778917dca8e35",
"version_minor": 0,
"version_major": 2
},
"text/plain": [
"interactive(children=(ToggleButtons(description='Plot Type:', icons=('',), options=(('Hyperparameters', 'param…"
]
},
"metadata": {
"tags": []
}
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "ATpPxL1spLjy",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "742531e4-a803-4243-ca65-cf1b9025f991"
},
"source": [
"plot_model(catboost, 'confusion_matrix')"
],
"execution_count": null,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"tags": []
}
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "X-7o23lc-bzA"
},
"source": [
"# Conclusions"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "UQTuQkQJKomt"
},
"source": [
"In our study we analyzed data from avalanches and measured weather variables for specific locations and dates to identify whether it was possible to predict an avalanche using the existing measured data.\n",
"\n",
"In this specific notebook the average of weather data from all locations was used in the analysis and prediction, so instead of focusing on particular data to understand avalanches in specific locations the study is more focused in a macro-weather environment that seems to be prone to avalanche incidence.\n",
"\n",
"By looking at the distributions of weather data from the days with and without avalanches we can see that Snow Water Equivalent is very important and this was confirmed when looking at feature importance in the final model. Also we were able to identify that the majority of avalanches happen in spring, probably because that is when snow is melting and mountains might be more unstable.\n",
"\n",
"The prediction model created is well capable to predict whether an avalanche might happen in one of the locations considering the overall weather data. However this model might not be good to predict the occurence of an avalanche in a specific location, as it uses that average of weather information. To get results for specific locations it would be recommended to collect weather data for as many locations as possible, and then cross those datasets to create the model."
]
}
]
}