Open source library
pymongo-3.11.0 ，matplotlib-3.1.3 ，
numpy-1.18.1，pandas-1.0.1，
scikit-learn-0.22.1，pyarc-1.0.25，
GDAL-3.1.2
Gets_the_location_where_power_consumption_occurs
Match the power consumption data and site location based on the site number.
Update_geographic_coordinates According to the address matching results of Gao DE map and Baidu Map, the number of distribution stations lacking geographical coordinates is reduced.
Geocoding_for_csv
The latitude and longitude coordinates of the distribution station are obtained through baidu map API according to the fuzzy place names of the distribution station, and the information of the power station with abnormal matching is given.
Normalization of training set data
Batch normalized data of each driver in the training set and record the parameters in the normalization process.
Regression_model_selection
By comparing the performance of different regression algorithms with the same set of data, the model most suitable for the experimental data was selected.
Get the best parameters
By controlling the variables, the optimal decision tree tree and the maximum characteristic number of the random forest model in the experimental scene were determined.
Establish the power consumption model
Based on the random forest algorithm, the power consumption distribution model is pre-trained and returns information such as the accuracy of the trained model and the importance ranking of the influencing factors.
Mapping of power consumption distribution
Based on the random forest algorithm, the power consumption distribution model is obtained through pre-training, and information such as the model accuracy of training and the importance ranking of impact factors is returned.
Chart the importance of drivers
According to the importance of driving factors in the model training process, draw the importance ranking of different regions
Plot the relationship between electricity consumption and GDP
Explore the correlation between power consumption and gross regional product, and draw an image