Using Semantic Features for Enhancing Car Pooling System

Nowadays the demand on carpooling system increases due to the need to decrease car crowdedness, saving fuel cost, decrease pollution, etc. Carpooling services depend on combining different passengers in one car who are willing to go to the same place in specific time. In this paper, a novel framework that utilizes trip profile and semantic of places (point of interest) is proposed. Users' trips are distinguished into routine trips and occasional trips. For occasional trips, the user is offered a similar destination based on the semantic of destination such that the new location is within accepted range or in the route with respect to drivers and other passengers. The proposed framework is applied on real dataset of New York taxi. Two techniques have been applied one based on route matching and other applied machine learning. The results show that the proposed framework outperforms tradition carpooling system by reducing total number of trips by 22.3 % in case of 3 passengers per car and by 26% in case of 4 passengers. While total number of trips has been reduced by 66% if 3 passengers accepted in one car, 74% if 4 passengers accepted when machine learning technique was applied on the same dataset.