The Assessment Of The Performance Of The Open-ended Funds Based On DEA-Neural Network

Data envelopment analysis (DEA) and neural network as a common method of evaluation of fund performance, has been widely used, but both methods have their shortcomings. DEA method is based on the theory of optimization based on linear programming theory, the interpretation of the real world will exist a large deviation, while the realization of DEA method is effective to a large extent by the number of variables, namely, the size of index system restrictions. As a kind of neural networks by actual output and the error between the target out put, continuous learning, revision, will eventually be reduced to a reasonable range of error to estimate the assessment methods, using artificial method to determine the designated target output with a high degree of subjectivity and randomness.Article selected early in 2006 to 2008 at the end of 74 open-end funds as the research sample, and selected to reflect open-end fund investment, risk and benefits of a total of 12 indicators in three areas involved in the evaluation, through data envelopment analysis, neural network 2 methods of model analysis, selection, constructed for this evaluation method. Through data envelopment analysis for the first time evaluation, and at the first evaluation results, based on the final evaluation of the use of neural networks, the conclusion that:(1) From 2006 to 2008, China’s open-end fund comprehensive inefficient Most of them are put into redundancy, lack of output, or a combination of both situations, which are non-optimal input-output accounts for the final evaluation of the fund 60% of the total number of samples. This phenomenon shows China’s open-end fund at a reasonable investment, as well as the utilization efficiency of the various elements of risk aversion, there are certain problems. At the same time from the empirical results found that the performance of bond funds than in equity and hybrid funds better, which can be debt hedge funds have better features to explain. (2) Data envelopment analysis and neural networks are two ways to apply the same open-end fund the comprehensive performance evaluation, you can better make up a simple method and simple to use DEA to use BP neural network deficiencies.