Context These datasets contain estimated speed and estimated trip durations for all trips generated by the fastest route dataset from oscarleo. The estimated trip durations are a better approximation for the actual trip duration than the total_travel_time variable from the fastest route dataset is. How these features are generated in detail can be found in this kernel: https://www.kaggle.com/pepeeee/nyc-estimating-avg-speed-using-fastest-route/notebook Content **estimated_speed** - This variable contains the estimated average speed of the respective trip. The unit of this variable is meter per second. **estimated_trip_duration** - This variable is generated by dividing the total_distance variable of the fastest route dataset by estimated_speed. The unit of the variable is seconds. Acknowledgements I'm very thankful to oscarleo for sharing the great dataset about the fastest routes. Without that dataset mine wouldn't have been possible. I would also like to thank saihttam for sharing the holidays package. Inspiration The fastest route contains a lot of very valuable data for predicting the trip durations of taxis in New York City. Just using the variables total_travel_time, total_distance and number_of_steps didn't feel like using its full potential to me.