Forecastor for Buildings' Consumption

Go check ce lien pour rédiger le README: Github-flavored Markdown

Installation

Command line: pip install building_energy_forecastor

Features' list for preprocessing data from src/building_preprocess

• day_of_week(date_serie): Takes a pandas.Series of dates and returns a pandas.Series of corresponding week days (['Monday', 'Tuesday', ...]).
• set_time_index(df, timeindex='Timestamp'): Set the time column as index of the dataframe df. By default the column's label is 'Timestamp'.
• time_to_cycle(df, timeindex='Timestamp'): From the 3rd competitor of the Forecast challenge by Schneider Electric. Add column to a copy of df containing cosinus and sinus functions of the time of the day, the month of the year and the day of the year.
• add_weather(df, weather, timeindex='Timestamp', freq_temp='D'): From the 3rd competitor of the Forecast challenge by Schneider Electric. Adds the weather data to the training dataset (df here) merging the two dataframes on the 'Timestamp' and rouding the time value in weather to the precised freq_temp ('D' by default).
• fill_temperature(df, tempindex='Temperature'): fill the NaN values in the tempindex column by computing the mean on the two closest framing values.

Model functions from src/building_model

• building_regressor(): Returns a linear regressor from Scikit-learn.
• building_train(reg, X, y): Trains the regressor with X the data and Y the targeted values.
• building_prediction(reg, X): Returns a pandas.DataFrame showing the prediction of the regressor reg given the data X.