Sales Volume and Price Prediction Using Attention-Based RW-FN-BiLSTM Framework

In the era of big data, accurate sales volume and price prediction play a crucial role in enterprises' market decision-making, inventory management, and pricing strategies. However, traditional forecasting methods often struggle to capture the complex temporal patterns and nonlinear relationships inherent in sales data. In contrast, deep learning techniques, particularly neural network models, have demonstrated significant potential in addressing these challenges. This study proposes a novel Attention-based RW-FN-BiLSTM hybrid neural network framework to enhance the accuracy of sales volume and price prediction. The model is trained and evaluated using a real-world dataset provided by the Hass Avocado Board from 2015 to 2023, covering multiple regions and key variables such as the price and sales volume of conventional and organic avocados. The proposed model integrates a Bidirectional Long Short-Term Memory (BiLSTM) network to capture long-term dependencies in time series data, Random Weight Functional Link Network (RW-FN) to enhance feature extraction capabilities, and an Attention mechanism to adaptively assign weights to critical features, thereby improving prediction performance. Experimental results indicate that the model exhibits stable convergence during training and testing, with predicted values closely following actual trends. Additionally, the error distribution is relatively symmetrical, demonstrating the model's strong generalization capability. However, some deviations remain, suggesting room for further optimization in network structure and feature selection. This study not only validates the effectiveness of the Attention-based RW-FN-BiLSTM model in sales volume and price prediction but also provides enterprises with a novel approach to market trend analysis, inventory optimization, and pricing decisions, enabling them to make more scientific and precise strategic choices in an increasingly competitive market environment.