PAM clustering based taxi hotspot detection for informed driving

A huge amount of spatio-temporal data is generated by millions of cabs in metro cities all around the world. This data, if analyzed correctly, can provide a better understanding of the taxi demand. With the increasing preference of customers to have a hassle-free experience, cabs are becoming the ultimate choice for all due to its point to point service. The inability of the taxi companies to meet ever increasing demand of cabs leads to high unavailability of cabs during peak hours and low usage during non peak hours. This taxi imbalance problem can be resolved by analyzing the spatial data and predicting the demand hotspots to identify areas with potential passengers. This paper aims at filtering the data based on the day of the week and the time of the day to segregate the data for further analysis. Clustering techniques are used on the filtered dataset to provide the popularity of regions in the city at different timings. K-means, k-medians and CLARA clustering techniques are used and the results from these clustering techniques are compared. The clustering technique providing the best results for spatial data is chosen for hotspot detection based on the day and time. This date and time based hotspot detection helps the taxi companies to dispatch the cabs at locations predicted to have higher number of passengers in future resulting in better service for customers and increased revenue for the shareholders.