Dynamic Pricing for New York Subway Fares

Inspiration

Some particular stations at NYC Subway are extremely overcrowded from time to time. Solving the overcrowding problem in subways is necessary for millions of commuters like us. Addressing this issue will lead to increased efficiency, reduced travel times, lower stress levels and decreased environmental impacts. NYC subway is crucial to the majority of the residents in the city. How to better serve the local community but also increase its ability to generate revenue is an important and fascinating problem.

What it does

Dynamic pricing model for the nyc subway fares which adjusts prices based on the real time demand of ridership.

How we built it

Used ARIMA time series model for predicting riderships which is then fed into our pricing model.Hybrid dynamic pricing model which assigns a price factor in alignment with the demand of ridership for any station.Clustering stations to ensure no nearby stations have very different pricing.

Challenges we ran into

The two datasets do not overlap in time so our original plan to also give a price to every pass fell short. We tried to merge different features from both of the datasets, but because of the enormous sizes of both datasets and having unequal number of instances led to some difficulties.

Accomplishments that we're proud of

Through simulation our model has proved to be successful in both reducing the overall price for locals and generating more revenue for MTA.

What we learned

We learned the importance of EDA in analyzing data and using a large dataset to fit into a big picture.

What's next for Dynamic Pricing for New York Subway Fares

Reinforcement learning models for strategy LSTM for time series forecast With complete data we can further try to assign a price for every subway product (full fare, 30 day unlimited, etc.)