Utilizing time series analysis to forecast the growth of mobile payment users and its implications for the digital economy

Abstract

Mobile payment systems have experienced rapid growth, but accurate forecasting remains challenging due to market dynamics and complex adoption factors. This paper proposes a Hybrid ARIMA-LSTM-Transformer model that combines time series forecasting, sequential learning, and attention mechanisms to address these challenges. Experimental results across five datasets demonstrate our model’s superior performance with MAE of 0.075, RMSE of 0.121, and R2 score of 0.948, outperforming traditional approaches. The model’s high accuracy and adaptability make it valuable for real-world applications in digital economy planning and mobile payment market analysis.