Vehicle Future Location Prediction Using Apache Spark for Optimal Performance

Abstract

The number of devices equipped with GPS sensors has increased enormously, which are generate a massive amount of data. Analysis can be carried out to use these big data in commercial and security related applications. One such application of these big data are to predict the future location of vehicles based on their previous locations. Using these predictions, different location-aware proactive systems can be develop. There are many models and algorithms which helps predict the future location with high probabilities. What is lacking in this field of study is the need of a system that utilizes minimum computing resources and can predict future locations with higher probability. In this study, we developed an algorithm according to the dataset used that results in lower latency and higher precision. Apache Spark, a big data platform was used for reducing the processing time and computing resources. The algorithm achieved 75% to 85% of accuracy and in some cases where the uses do not change their daily routine frequently to 100% accuracy for all previously visited locations. We compared the prediction results of Apache Spark with simple python algorithm without using Apache Spark and experimentally found that Apache Spark predicts processes up to 300% times faster. This algorithm can help find useful knowledge for commercial, intelligence and other security related application.