Analyzing the Accuracy of Clustering Based Cab Recommender System (CBCRS) Across Varying Densities

Abstract

Recommender systems are widely used in today’s world as they narrow down a huge number of options for a user in various fields. These recommender systems can also be significantly used by cab drivers also, since they can recommend the nearest location of passengers waiting for a cab. For the system to perform better, cluster formation is an essential step to be performed using the pick-up locations of the cab provided in the historical data. The recommendation is highly dependent on various other parameters also like timestamp, density of the area, clustering technique used, etc. This research paper aims to analyze the impact of varying minimum density of the pick-up data points on the accuracy of the recommendations by Clustering Based Cab Recommender System (CBCRS). Therefore, the research paper has a threefold objective: Firstly, to identify different clustering techniques for clustering the pickup locations. Secondly, the research paper proposes the design and development of a framework for CBCRS to cluster the cab pick-up data points along with recommending the nearest pick-up location of passengers to the cab drivers. Finally, the paper analyzes and evaluates the accuracy of CBCRS recommendation across varying densities of the pick-up data points as compared to the Partition-based Clustering Method, Density-based Clustering Method and Hierarchical Clustering Methods.

Keywords

• clustering
• intrinsic parameters
• cluster evaluation
• recommender system
• unsupervised cab dataset