ARIMA, Prophet, and LSTM-based analysis of demographic factors in smartphone usage patterns

Abstract

In today’s digital era, the threat of problematic smartphone usage is very prevalent. To mitigate this threat, a deeper understanding of user behavior is essential. This study focuses on the prediction of problematic smartphone usage patterns among users, considering various demographic variables (gender, marital status, employment, and education). To achieve the study aims, the WhatsApp status seen time primary data is collected from 189 participants for 128 days from Indian students representing different demographic backgrounds. To analyze the collected data, we employed descriptive statistics with three prominent time series models, namely ARIMA, Prophet, and LSTM. The results posit that females, bachelor’s degree students, unmarried, and unemployed participants were found to have a relatively higher risk of problematic smartphone usage. Lastly, the results confirmed that the ARIMA forecasting algorithm is more efficient in forecasting behavior than Prophet and LSTM. While the prophecy algorithm gives better results than LSTM. To the best of our knowledge, none of the previous studies considered marital status and employment status as analysis parameters, and no study used time-series data to provide insight into problematic smartphone usage. The study findings can prove to be a better guide for parents, psychologists, educators, social workers, and policymakers in understanding problematic smartphone usage among students, who are the youth and future of the country.