Abstract:
In recent times location prediction has received a great interest in research due to its application in diverse areas. Its usage is found in courier firms, taxi companies, detective organizations and advert placement firms. These organizations are challenged with some issue of concerns such as not been able to capture semantic information and also keeping track of mobile users movement. This work focuses on the use of Spatio- Temporal Recurrent Neural Network (ST-RNN) and a machine learning model known as the Markov Model (MM). This research attempts to provide solutions to the obvious challenges of past works by developing an efficient and robust model for location prediction. Sequel to the adoption of the aforementioned techniques, preprocessing of the dataset was done by removing the irrelevant features from the datasets, separating the time and the date that were initially merged and formatting the longitude and latitude columns into the required format. Consequently this system combines the strength of the two technique for predicting user’s location; the experimental results yielded a high level of accuracy, took less computational time as well as a reduced system error rate. System evaluation using computational time, root mean square error, prediction accuracy and mean square error of the system when compared with other state of the art technique revealed that combining ST-RNN with Markov Model produced a model with a higher level of accuracy.
