Digital Economy Market Forecasting Model Based on Deep Learning Algorithm

This paper aims to solve the problem of insufficient data processing capabilities of traditional methods in digital economy market forecasting and the difficulty in modeling nonlinear and time series relationships. Combined with the advantages of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) Network, high-dimensional data features and long-term dependencies are fully explored, improving the accuracy and reliability of e-commerce product sales forecasting in the market. To ensure the high quality and relevance of the original input data, this paper performs data preprocessing steps such as data cleaning, outlier processing, and missing value processing on the original data. The model training is based on the number of clicks, views, conversion rate, sales volume, sales volume, etc., of e-commerce products. CNN is utilized to extract features from multidimensional data to capture complex local features and key features. LSTM is used to model long-term dependencies in time series to ensure that the model can effectively handle dynamic changes and trend predictions. The outcomes demonstrate that the CNN-LSTM composite model is superior to the traditional ARIMA model in many evaluation indicators. The MSE and RMSE values are reduced by 72.2% and 47.0% respectively compared with the traditional ARIMA model, and the MAPE value is further reduced to 1.39%. The R2 value of the composite model reaches 0.97, and it performs well in real-time indicators such as throughput and delay. The CNN-LSTM composite model studied in this paper provides an efficient and precise solution for digital economy market forecasting, and provides more scientific and effective decision support for enterprises and decision makers in the digital economy.