Return Rate Prediction Model Using Traitor Feline Crow-Based Hybrid Long Short-Term Memory and Light Gradient-Boosting Machine Model

Return rate prediction involves forecasting the rate at which products or investments are returned, driven by factors such as customer dissatisfaction or financial performance. This predictive capability is crucial for businesses and financial institutions, as it facilitates improved decision-making, optimized inventory management, and enhanced risk assessments. However, existing predictive models are often constrained by their inability to fully capture complex, sequential patterns in data, their limited capacity to handle both temporal and non-temporal features effectively, and the challenges of balancing predictive accuracy with computational efficiency. To address these limitations, this research introduces the Traitor-Feline Crow Optimization-based Hybrid Long Short-Term Memory and Boosted Gradient Boosting Machine (TFC-LSTM boosted GBM) model for return rate prediction. The proposed TFC-LSTM boosted GBM framework excels in capturing sequential patterns and temporal dependencies, leveraging historical data trends to enhance predictive accuracy. The model strategically optimizes data utilization, effectively reducing prediction errors and improving overall performance. By employing adaptive strategies, the TFC-LSTM boosted GBM framework navigates diverse data landscapes with precision and intelligence, seamlessly integrating temporal data handling with efficient tabular data processing to create a robust predictive framework. Experimental results validate the efficacy of the proposed approach, demonstrating its superior performance using the Bitcoin price prediction dataset. The model achieves exceptionally low error rates, with a Mean Absolute Error (MAE) of 1.31 and a Mean Absolute Percentage Error (MAPE) of 3.48, underscoring its potential as a reliable and efficient solution for accurate return rate prediction.