A Prediction Model for Human Happiness Using Machine Learning Techniques

Abstract

Organizations move forward when employees feel fulfilled and happy at work [1]. Thus, most organizations need to investigate employee attitudes toward work [2]. One effective way to obtain employee information is by survey. The dataset analyzed in the research was collected by questionnaire in 2017. The respondents were employees who worked for the Ministry of Public Health in Thailand. A happiness prediction model was created based on the information from the survey. The experiment employed four machine learning techniques, namely KNN, Decision Tree, Naïve Bayes, and Multi-Layer Perceptron. Two major techniques, namely over-sampling and under-sampling, were also adopted in the research to maintain a class distribution. The experimental results show that when an imbalanced data issue was solved, the prediction accuracy was higher than the results from the original data. The prediction accuracy of the proposed model was approximately 87.66 percent.