Power-Outage-Prediction

Predicting power outages using extreme weather events

This project is structured as follows.

• codes contains the python notebooks for data preprocessing, exploratory data analysis, and modeling.
  • codes/preprocessing contains the python notebook merge.ipynb, where we read power and weather data and merge them together into one dataset, indicating whether a weather event yielded a power outage.
  • codes/eda contains python notebooks for exploratory data analysis, one for each type of weather.
  • codes/modeling contains python notebooks for training models and evaluating their performance. For each weather type of interest, there is one notebook including brief, organized code illustrating how we train models for the given type of weather. Additionally, there are subdirectories for each weather type, containing the more detailed used during our experiments.
  • codes/results contains python notebooks describing the results of our trained models.
  • codes/misc_old contains miscellaneous/old code not used in the final version of our project (kept for posterity).
• figures contains different figures, e.g, confusion matrix, roc curve, important features etc, which summarizes all the models for all three weather types.
• merged contains the merged datasets outputted by merge.ipynb
• misc_old contains miscellaneous/old files not used in the final version of our project (kept for posterity).
• models contains the saved, trained models for all three weather types.
• power_data contains the datasets for power outages.
• weather_data/cleaned/ contains the cleaned weather data for different weather types.