R&d Project Decision Of Wavelet Neural Network Research

In the background of increasing degree of globalization, Technology plays a significant role on society, economic progress, improving the quality of life, and even the enhancement of comprehensive national strength. With the scarcity of scientific and technological resources in China’s enterprises, how to effectively use resources to the real prospects for the development of projects, has significant meaning for improving the success rate of R&D projects, thus enhance the competitiveness of enterprises.R&D project has the largest characterize of risk and uncertainty, in this case, how to terminate an unnecessary R&D project in the appropriate time to reduce the loss of business is a worth exploring question. Because failure R&D project often drag on for a long time but no avail, it will be virtually bound to cause tremendous losses to the enterprise.This paper is based on management practices, combines with the suspension of the decision-making about the status quo at home and abroad, it starts with the meaning of R&D project termination decisions, to analyze the need of termination decision.Wavelet neural network method of R&D project termination decision is a new type of neural network based on wavelet analysis, It is a good combination of self-learning features from BP neural network and time-frequency analysis features from wavelet analysis.Then, based on phased characteristics R&D projects and sources of risk, risk characteristics, risk distribution, risk responses, it combines with the domestic and international factors set of indicators, proposes the principles of indicators establishment and establishes systematic indicators.Through the establishment of relevant indicators, BP neural networks and wavelet neural networks will be applied to the R&D project termination decision, and through analysis of the two instances of application, the two models will be compared.After instance analysis, It will be confirmed that wavelet neural network method is more effective when compared with the traditional BP neural network. And it improves the lower status when faced with few sample data. It will prove the validity of construction of wavelet neural network model.