Attention Mechanism-Based CNN-BiLSTM for Sales Revenue Prediction

In the era of the Internet, the development of precise sales forecasting models is crucial for assisting businesses in adapting to market demand fluctuations and formulating rational inventory plans. Due to the influence of numerous nonlinear factors, the accuracy of traditional sales forecasting models often falls short of satisfaction. To enhance the precision of sales forecasting, this paper constructs an attention mechanism-based CNN-BiLSTM (Convolutional Neural Network-Bidirectional Long Short-Term Memory) sales forecasting model using historical entity data. Comparative experiments are designed and implemented, involving a single model and two typical time series forecasting models. The experimental results indicate that, under identical experimental conditions for enterprises, the attention mechanism-based CNN-BiLSTM forecasting model, compared to other models in predicting product sales revenue, demonstrates superiority. The findings suggest that the CNN-BiLSTM network model based on attention mechanism excels in forecasting order sales revenue, offering a more accurate prediction of future sales.