Success Rate Prediction of Instant Carpooling Based on Random Forest

Currently taxi apps have been popular in life and carpooling has become a new trend of saving money and environmental protection. However, during carpooling, especially in an instant carpooling scenario, it may often fail because the system may not be able to match a suitable co-passenger in a short time, resulting in direct abandonment or missed potential chance of successful carpooling. A model of instant carpooling prediction based on random forest is proposed, its generalization ability and the prediction performance are greatly improved through feature extraction and feature derivation. The model is validated through the analysis of a real order dataset and proved better performance than other classification algorithms.