Research On The Application Of Averaged One-dependence Estimators Algorithm In Personal Credit Evaluation

With the rapid development of Chinese market economy,personal credit as a guarantee to bank loans has become a trend.As both the number and amount of credit loan increase,the risks caused by the customer's defaults is also increasing year by year.As a result,before offering credit loans to customers,it is necessary for the bank to assess the credit objectively and accurately according to the customer's real information and then accordingly provide the loan,thereby reducing the economic loss to the bank due to the customer's breach of contract.Based on the careful analysis of the continuous attributes in the sample data set and characteristics of the AODE(Averaged One-Dependence Estimators)model,Firstly,the discretization method is proposed in this dissertation.after discretizing the continuous attributes,the classification accuracy of the AODE model can be improved effectively;then,on the condition that the attribute reduction rule of the rough set theory is capable of keeping the classification capacity of the information system unchanged,the irrelevant or unimportant index attribute is deleted so as to select the minimum index set that can represent the sample;next,the Adaboost is combined with the AODE model to form the integrated AODE classifier;finally,the sample data are input to build the personal credit evaluation model.The main work of this dissertation is as follows:First of all,in order to solve the problem of discretization of continuous attributes in a data set,one improved discrete particle swarm optimization algorithm is proposed.Taking the breakpoint set of the continuous attribute as the discrete particle swarm,the breakpoint subset is minimized through the interaction between particles.At the same time,the simulated annealing algorithm is introduced as the local search strategy for improving the diversity of particle swarm and the ability to find the global optimal solution.In order to achieve the purpose of discretization of continuous attributes,the consistency of decision table is determined by the dependency of decision attributes on the conditional attributes in rough set theory.Secondly,as most index attributes in the sample set are redundant and do not have the same importance,which is not conducive to make a concise decision in data analysis,the reduction of index attribute in the sample data set is a vital step for credit assessment.The attribute reduction algorithm based on the tabu discrete particle swarm optimization is put forward for the selection of personal credit evaluation index.Since the tabu search algorithm has a strong dependence on the initial solution while the discrete particle swarm algorithm is easy to fall into the local optimal solution in the iteration,the discrete particle swarm algorithm is used to search the global optimal solution and the tabu search algorithm is used to search the local one when selecting index with the attribute reduction.Without affectingthe quality of classification,the redundant attribute is deleted to simplify the knowledge base,building the set of personal credit evaluation index.Thirdly,although each Super Parent One-Dependence Estimator(SPODE)model contributes the same to the classification when AODE model is used for classifying,their structures as well as the effects on the final classification are different.In terms of the structural weakness of the AODE model,the improvement measures are proposed in this dissertation.First,among the SPODE models composed of the AODE model,a certain number of SPODE models are selected to form the AODE model using the random sampling method;then the Adaboost algorithm is adopted to construct the integrated classification model;Finally,the integrated AODE classification model is used for personal credit evaluation.The simulation experiment results show that the integrated AODE evaluation model can effectively improve the prediction accuracy of personal credit evaluation.