Travel Time Prediction in Ride-Sourcing Networks : A Case Study for Machine Learning Applications

This paper explores the applications of machine learning for predicting the travel time in the ride-sourcing networks using the Uber movement dataset. Using the Python programming environment, a case study is presented to analyze the travel time of the ride-sourcing services from the central Washington D.C. to the given specific destinations by considering the distance, railway/subway and street density in different destination zones (areas) and also weather conditions. To this end, in the first step, a descriptive analytics is completed to include potential features (attributes) affecting the travel times of Uber (ride-sourcing) services. Then, machine learning techniques such as random forest and robust regressions are applied to identify key attributes (features) for the prediction of the average travel times. The findings and accuracy of the robust regression models are compared with the random forest to select the best model in predicting the mean travel time. This case study provides opportunities in data preparation, descriptive and predictive analytic topics covered in applied machine learning, data science and decision support system courses using data mining programming environments like Python and R. Students are also able to change the study area (city) for this case study based on their interest