We have a year of data for each time a customer has requested for a ride. Can we try and predict ride requests for the future?

The data is stored in this format - timestamp, customer_id, pickup request latitude, pickup request longitude, drop request latitude, drop request longitude. A snapshot of the data is below.

There were 8122696 ride requests recorded.

• These 8122696 observations were recoreded over a span of 363 days.

• From the data, it was clearly visible that both the trend and seasonality was present in the data.

• The time series is not stationary.

• Clear spike in the data at every seven day interval

• Based on the forecasted data, the marketing strategy can be planned for the upcoming months.

• Increasing pattern in the ride request is seen in the forecasted data.
• Forecasted ride request during the weekday is much more than the weekends.

An abrupt change in the trajectory of the time series is usually seen once in the middle and once towards the end of almost every month.

This analysis was done on the entire dataset, however this can be further broken down based on locations and different points in times during the days.