DEVELOPING A MODEL FOR MULTILEVEL STRESS PREDICTION USING DATA MINING TECHNIQUE

Abstract

This research aims to study and find the optimal stress prediction model based on data mining techniques. In the research procedures, we let 300 sample testers do a self-analysis stress test 4 times, while each time has been done every two months. Next, these raw data were used as the input data sets of six data mining algorithms based on WEKA to develop the prediction models. In this research, we studied six algorithms, i.e., Bayesian Network, Naïve Bayesian, Decision Tree:4.5, Decision Table, Partial Rules (PART), and Multilayer Perceptron (MLP). To evaluate these six models, 10-fold cross-validation were utilized to split the data into a training set and a test set. The experimental results show that the MPL algorithm with the data set of 3 test times during the last 6 month-period is the optimal prediction model comparing to the other five models. That is, its accuracy, precision, recall and F-measure are 81%, 0.81, 0.81 and 0.81, respectively