Research on Product Demand Forecasting Based on Random Forest and ARIMA Time Series: Precision Forecasting Method for Data-Scarce Environments

Abstract

In the field of e-commerce, accurately predicting the future demand for specific products is crucial for efficient inventory management and supply chain optimization. Although traditional time series forecasting methods have shown reliability in handling rich data, they are significantly affected by seasonal fluctuations and sudden market events. This is especially true for newly launched products or long-tail products with insufficient sales data, which often results in poor performance, limiting the accuracy and feasibility of traditional models in practical applications. To address this issue, this paper proposes an innovative forecasting framework based on extensive historical sales data from e-commerce platforms. This framework employs random forests to deeply mine and learn from historical data, and uses the K-Means clustering algorithm and Euclidean distance to identify the most matching sales feature trends for target products with limited data. It further provides accurate predictions for these products through the ARIMA time series model, offering robust support for further optimization of the forecasting model. Ultimately, by comparing with traditional models, the innovative forecasting framework presented in this paper demonstrates a significant improvement in accuracy when predicting sales of products with limited data, overcoming the challenges of forecasting under data constraints, and fully proving its effectiveness and innovativeness.