Performance Analysis of the IndoBERT–Prophet Hybrid Model for Logistics Applications

The increasing competition in Indonesia’s logistics sector, particularly in digital courier applications, highlights the need for advanced analytical tools capable of understanding and predicting customer sentiment in real time. However, current sentiment analysis methods often lack contextual depth and predictive capability, limiting their practical value for decision-making. This study aims to develop and validate an integrated analytical framework that combines diagnostic and predictive analytics for logistics performance evaluation. The framework integrates a fine-tuned IndoBERT model for sentiment classification and a Prophet model for time-series forecasting, allowing the analysis of user reviews while accounting for external service disruptions. Empirical validation was conducted using 924 user reviews from the PosAja! application by PT Pos Indonesia. The IndoBERT model achieved an impressive 99.5% accuracy, effectively identifying two main complaint categories: application functionality issues and delivery delays. The Prophet forecasting component successfully modeled sentiment trends, revealing spikes in negative sentiment that strongly correlated with technical service disruptions, such as COD feature failures and server maintenance. The results confirm the framework’s robustness in both diagnosing and forecasting sentiment dynamics. User sentiment proved to be a sensitive real-time indicator of service stability and operational performance. The validated IndoBERT–Prophet hybrid framework provides a novel, data-driven approach for proactive decision-making and continuous service improvement in the logistics industry.