Evaluating predictive analytics model performance accuracy for network selection mechanism

Abstract

Predictive analytics has been widely used and adopted in many fields. The idea of anticipating change rather than reacting to change has appealed to many system designers. In this paper, we evaluate the feasibility of applying predictive model into a network selection mechanism to choose most reliable network with higher speed for a communication device such as a modem. The predictive model will attempt to predict best network download speed at a given time of day based on the historical data that we have measured at specific location under studies. This paper also focuses on the accuracy of these predictive models on our sample data, which are measured by calculating its Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) values. Based on the results, a decision tree model outperforms linear regression and M5 models in terms of accuracy. The findings of these studies help us to improve our cognitive network selection algorithm in making decision for best network.

Keywords: Network Selection Mechanism, Predictive Models, Linear Regression, Decision Trees, M5