Transformative computing for products sales forecast based on SCIM

Abstract Online agricultural product trading has the characteristics of rapid and diversified transaction data; there is a fuzzy correspondence between sales volume influencing factors and sales volume levels. Based on this, this paper combines the data preprocessing technology of fuzzy membership and optimized deep learning algorithm, adding a self-encoding method with sparseness restriction, and proposes a deep learning sales forecasting model based on transformative computing with fuzzy membership-the super crown model (Super Imperial Crown Model, referred to as SICM). The model uses fuzzy membership to process the weighted relationship between sales influencing factors and sales rank, and uses a sparse autoencoder network to adaptively extract sample features; sales rank classification prediction uses Softmax classifier; BP fine-tuning is used to Achieve parameter optimization. Finally, use the collected transaction data to apply R software to simulate the optimized model and compare and analyze the comprehensive prediction performance. The results show that the super crown model can realize real-time and accurate dynamic sales classification prediction according to the characteristics of current online agricultural product transaction data, effectively overcome the imbalance of supply and demand caused by information imbalance, and promote the study of deep learning in the field of e-commerce transactions effect. Presented algorithm based on transformative computing techniques can be used in optimization of sales processes, management and analysis of sales markets.