Time Series Prediction using Spark

These days in high-tech or smart cities the pedestrian counts can be monitored by deploying sensors at certain locations which can count the number of pedestrians every hour(as per the data used for this blog) or as required. From the title of the post itself one can understand that here we will try to predict the count of pedestrians or pedestrian traffic at certain locations for the next hour from the data of previous hour(s). This technique is also called a one-step time-series prediction, where we are predicting the next value with the previous values. Therefore this is a time-series regression type of problem as the data to predict is of continuous nature. Using these predictions we can select the locations with most traffic which can then be used by certain companies to market their products, performers in the music and entertainment industry to make sure that they are heard by the most amount of people, etc.