Analysis of the Happiness Level of Indonesian Society Using the Long Short-Term Memory (LSTM) Model

Abstract

Happiness is an important indicator for assessing societal well-being. Various research methods have been employed to analyze people's happiness levels; however, studies utilizing soft computing approaches, including Long Short-Term Memory (LSTM), remain limited. This study aims to analyze the application of the LSTM model and describe its performance in predicting the happiness levels of the Indonesian population. The data were obtained from the fifth wave of the Indonesian Family Life Survey (IFLS), comprising 1,642 respondents. The independent variables included age, marital status, gender, employment status, highest educational attainment, life satisfaction, economic status, health, religious devotion, hope, and personality, while the dependent variable was happiness, which was classified into four levels. The LSTM model used to predict the happiness levels of the Indonesian population employed three gates, namely the forget gate, input gate, and output gate, with four units in the LSTM layer and 10 training epochs. The results showed that the model achieved an accuracy of 77%, precision of 77%, recall of 100%, and an F1-score of 87%. These findings indicate that the LSTM model performs well in predicting the happiness levels of the Indonesian population. The performance of the LSTM model is also influenced by the selection of independent variables. Therefore, the findings further suggest that the independent variables used in this study may serve as meaningful indicators of happiness among the Indonesian population.