Chapter 12: Predicting Taxi Demand on the Spark Platform

Demand prediction is one of the most popular applications of time series analysis. We can predict, for example, the demand for electricity in households, restock in the retail industry, and taxi drives in a large city. Regardless of the application, the idea is the same: use historical data and possibly some external information to predict the future demand. Then, use the predictions to optimize the supply chain or service management. What varies between the applications is the length of the forecast horizon and the granularity of the historical data. While restocks might be planned for the upcoming months based on daily data, the size of a taxi fleet might be adjusted for the next days or even hours, based on hourly data.

Therefore, different demand prediction applications work on very different data volumes. Historical data that adds up every hour or minute will likely result in a much larger volume than data that updates...