Research And Application Of Sales Forecast Based On Neutral Network

In the worldwide, CRM (Customer Relation Management) has been greatly developed since 1997, and it becomes an important part for enterprise information infrastructure. In the same time, more and more customers' data have been accumulated. As the competition between companies becomes fiercer, companies have advanced requirements on decisions supporting. The combination of customer relation management and data mining is a trend now. This paper applies data mining technology to sales forecast module to meet the demands of decisions supporting in CRM.Structured CRM system's design based on modules, functions among different modules are isolate. Data interaction process is complicated. This paper applies BPM method to CRM system design. By completing business process, system provides different functions, and management staff can manage and optimize the business processes. This solution effectively combines different information systems within company, provides convenience for building data warehouse and data mining.Statistics based sales forecast methods often just consider one part of elements that affect sales, it cannot indicate inter-affection between all elements and the model is simple. By comparing statistical methods for sales forecasting and artificial neutral network, we can find ANN method has better non-linear mapping and self-learning ability. It can solidify all the factors and the relations between factors into the sales forecasting model through learning.This paper gives a detail analysis on sales forecasting based on back propagation network. That includes 3-layer BP network forecast model, network structure, the number of hidden nodes, choose and pretreatment of sample data and determination of preliminary parameters. The prediction results indicate that the sales forecast precision is improved, the capability of generalization is good.In the forecast system which initialized and developed in this paper, is applied to sales analysis and forecast for mid-small size company. It bases on different information systems within a company to build data warehouse, using data mining for sales forecast. And also includes other data analysis models. Considered system designing and structure, in order to integrate other types of neutral network application to this framework in future, this paper extracts common characteristics and designs interfaces for different neutral network, and this provides support for future functions enhancements.