Research On The Performance Evaluation Of Enterprise Knowledge Management Based On Improved BP Neural Network

In the era of knowledge economy, knowledge contained in human capital and technologyis an important capital for enterprise to win in the fierce market competition. Thecontemporary enterprises operating model has mainly converted to the development of humanand intellectual capital, knowledge management will become the focus of enterprisemanagement. The efficiency of enterprise knowledge management activities depends onwhether it could implement performance evaluation and whether the results of evaluationcould produce effective feedback and improvement to enterprise knowledge managementprocess. Through the evaluation of the status of enterprise knowledge management,organizers could discover the problems in the enterprise knowledge management, and thenrationally allocate corporate resources. Knowledge management performance evaluation is acomplex process related to many factors, because reliable solutions based on accurateproblems, reasonable and effective evaluation model is needed to realize the accuracy.Firstly, combined with the actual situation and relevant literature, the paper builds a twoevaluation index system. According to the index weight by AHP method,8representativeindicators are selected as the evaluation indexes. Then uses BP network to evaluate theperformance of knowledge management, designs network parameters and builds networkmodel. Dividing the data into training set and test set to simulate, the simulation results are farfrom ideal, then analyzes the cause of low accuracy, proposes using improved BP neuralnetwork algorithm to train the network, which includes the momentum-adaptive learning ratealgorithm and LM algorithm. According to the simulation results of the three improvedalgorithms, analyses the superiority of improved algorithm respecting to the standardalgorithm. Comparing the performance of three improved algorithm, momentum-adaptivelearning rate algorithm is proposed as the final training algorithm. Analyzing the simulationresults and the results of expert evaluation, we find that the improved BP model is feasibleand accurate. Finally, summarizes the full text and points out the direction of improvement.