Principal Component Fuzzy Neural Network In The Research And Application Of The Scientific Research Innovation Ability Evaluation In Colleges And Universities

During the course of building an innovative country and enhancing the independent innovation capability, universities are the main force and the important source of high-tech innovation. The evaluation on the university’s innovation ability, not only may improve university’s efficiency and level of scientific research, but also make a significant sense to perfect the china’ scientific research innovation system.Based on Referring to the recent research achievements at home and abroad, research and design work was carried out in the following area:(1)The paper designs the multi-university research innovation ability evaluating indicator system.By the principle of science and justice, through questionnaires, expert opinion and reference to relevant research results, the paper designed the multi-university’s research innovation ability evaluating indicator system.(2)The paper studies a variety of typical evaluation models and methods.The paper studied many typical evaluation models and methods. And focused on studying these evaluation methods, such as Principal Component Analysis(PCA),BP Neural Network, Takagi-Surgeon(T-S) model based Fuzzy Neural Network(FNN) and SOM Neural Network. Though comprehensive analysis, two evaluation models was put forward. One was based on the Principal Component Analysis and BP Neural Network(PCA-BP).The other was based on the the Principal Component Analysis and Fuzzy Neural Network(PCA-FNN).(3)The paper finishes evaluating the university’s scientific innovation ability by using the PCA-BP model and the PCA-FNN model respectively.In the principle of minimum data information loss, the paper first uses Principle Component Analysis (PCA) to reduce the dimensionality effectively and filter out the main factors about90%, which aimed at eliminating the data’s correlation and overlap. Then the paper constructed the evaluation models based on PCA-BP and PCA-FNN respectively, and simulated,checked,test and completed the evaluation of the university’s innovation ability with Matlab.At last the better experimental results of PCA-FNN was not ranked directly, but grouped by cluster. The result was remarkable. Using this cluster method, we can eliminate the harm by direct scoring rank.(4)The paper compares two evaluation models between PCA-BP and PCA-FNN.Compared from data processing, training times, goodness of fit, evaluation result, the paper drew a conclusion that the model of PCA-FNN was better in goodness of fit and accuracy. And it was feasible.The research and application of PCA-FNN is proved a new method and made a significant attempt for the university’s evaluation of research innovation ability.