Designing a Predictive Model for Train Arrival Time Management, Using Data Mining Approach

Abstract

Access to transport service is critical to the development of all aspects of a nation train arrival time management including staff behavior, affordability, and ticket payment system and also somewhat satisfied with reliability, comfort, safety and security accessibility and availability. However, this transport services are not free from problems. Passenger loading is the main problems of all railway services operators. This research therefore aims to design a predictive model that can determine Train Arrival Time Management of Addis Ababa light transit operating control center data. To overcome the drawback of simple statistical method, we proposed the use of data mining techniques, for the data analysis for train arrival time management. The study follows hybrid data mining process model. After experiment survey for problem understanding, selected around 20,000 records of three years from OCC data. After eliminating irrelevant and unnecessary data, a total of 15040 datasets with 12 attributes are used for the purpose of conducting this study. Data preprocessing was done to clean the datasets. After data preprocessing, the collected data has been prepared in arff format suitable for the DM tasks. The study was conducted using WEKA software version 3.8 and three classification techniques; namely, J48 algorithm from decision tree, Naïve Bayes and JRIP, algorithm from rule induction. As a result, J48 decision tree algorithm with Percentage split (66%) registered better performance of 95.5612% accuracy. As a result, the study showed that scoring high value in speed, headway time and passenger loading attributes in train arrival time management are determinant factors for the arrival time success in the AALRT. Besides, the study revealed that other regions train arrival time management is more associated with success rate.