Prediction of hourly rental sharing bicycle

This repository contains the source code of the Prediction model for hourly bicycle utilization.
Details on the hour.csv and day.csv datasets can be found on Readme.txt.
The Jupyter notebooks contain the corresponding files for the model development. It is organized sequentially as follows:

1. Explorative Data Analysis
2. Model Development (Feature Engineering, Scaling, Split train-test sets), Forward cross-validation
3. 3.1 Model error (R2, MAE, MAD) using test set, feature importance
   3.2 Export model for production, recreating a client app to conduct inference via HTTP request
4. Two extra models:
   • A Linear Regressor using a QuartileTransformer for the target
   • A Neural Network Regressor (Multi-Layer Perceptron Regressor)

To see the model error, and conduct inference of the exported model check notebooks 3 and 4