Researches On Sub-contractor’s Qualification Comprehensive Evaluation Method Based On BP Neural Network

This paper analyzes the current development outlook and competitive advantageof Chinese telecommunication enterprises, focusing on the widely-appliedsub-contracting strategy during projects delivering, we find that it is very importantand necessary to launch a comprehensive evaluation on vendors’ qualification. Basedon the characteristics of telecommunication project delivery, the paper splits thefactors which affect the vendors’ qualification, then build a vendors’ qualificationindicator analysis model, so we can analyze vendors’ qualification comprehensiveevaluation system, this paper mainly discuss the contents below:Firstly, make a research on the characteristic of telecommunication projectdelivering and evaluation methods, additionally a project characteristic comparisonwas made on domestic and abroad. Then a vendors’ qualification analysis modelcan be built based on them.Furthermore, analyze the principle of building a vendors’ qualification figuresystem using Fishbone, Man-device-environment, AHP analysis methods, finallybuild a vendors’ qualification comprehensive evaluation system with4basic factorsand35figure factors.Then briefly research the classic algorithm of neural network, regarding to thedefect that BP neural network may easily sink into partial minimum during training,we use the extra momentum algorithm to solve it, further build a vendors’qualification comprehensive evaluation model based on the design of network layers.Last, verify the accuracy that BP neural network applied to vendors qualificationevaluation, we prepare25sets of sub-contractor qualification samples and train thenetwork, we find the possibility that BP neural network may converge slowly or notconverge for a long time, can be improved by using new activating function andself-adapted learning rate algorithm. Then simulate the model using MATLABtoolbox, compare with expert evaluation, results proves that the network establishedby computer can reflect the complex, non-linear relationship of vendors’ qualificationfactors, and further realize the function of prediction.