Predicted number of pickups, given location cordinates(latitude and longitude) and time of the day.
Time-series forecasting and Regression.
New York citymap is divided into several regions based on region radius and density of passenger.
K-means clustering used for geometric division of the city map
Time series data is divided into 10 minutes interval.
Frequency domain features are also used along with time domain data to improve the model.
Linear Regression, Random Forest and Gradient Boosted Decision trees(GBDT) models were trained and tested.
GBDT model performed better than other models