Taxi travel time prediction using ensemble-based random forest and gradient boosting model

Proposed work uses big data analysis and machine learning approach to accurately predict the taxi travel time for a trip based on its partial trajectory. To achieve the target, ensemble learning approach is used appropriately. Large dataset used in this work consists of 1.7 million trips by 442 taxis in Porto over a year. Significant features are extracted from the dataset, and Random Forest as well as Gradient Boosting is trained on those features and their performance is evaluated. We compared the results and checked the efficiency of both in this regard. Moreover, data inferences are done for trip time distribution, taxi demand distribution, most traversed area, and trip length distribution. Based on statistics, errors, graphs, and results, it is observed that both the methods predict time efficiently, but Gradient Boosting is slightly better than Random Forest.