Sales Prediction in Omnichannel Distribution Systems Based on Consumer Behavior Using Process Mining and ARIMAX

Omnichannel distribution integrates multiple sales channels into a single management platform to enhance customer experience. However, the complexity of consumer behavior data across channels often prevents business owners from effectively analyzing and predicting sales. This study proposes a sales prediction model in omnichannel distribution systems based on consumer behavior using a collaborative approach that integrates process mining and autoregressive integrated moving average with exogenous variables (ARIMAX). Event log data from an omnichannel service provider were used to extract consumer activity patterns, which were then analyzed through process discovery algorithms to identify dominant behavioral processes. The resulting behavioral indicators served as exogenous variables in the ARIMAX model for sales forecasting. The experimental results show that combining consumer behavioral data with ARIMAX improves prediction accuracy, achieving a mean absolute percentage error (MAPE) of 2.5% after logarithmic transformation. The findings demonstrate that consumer behavior significantly contributes to improving sales prediction accuracy, providing valuable insights for business decision‐making in omnichannel environments.