For BA-FRVM Research And The Quantitative Prediction For Typical Faults Parts In Automobile

In recent years,with the rapid development of automobile enterprises and the surge in the volume of data in the industry,automobile companies urgently need to find the pattern from the data in order to predict accessories needs and diagnose car fault.Consequently,many artificial intelligence algorithms are used in the automotive industry.The commonly used artificial intelligence algorithms are support vector machines and BP neural networks.However,their shortcomings inhibit their applications.In this paper,we use BA-RVM optimized BA-FRVM algorithm for the number forcast of vehicle failure parts.In this paper,a forecasting model for typical fault pieces of automobile based on the fast relevance vector machine of bat algorithm is established.All simulation experiments are implemented by using Matlab R2014a software.Firstly,the factor-kernel parameters,which significantly influence the accuracy of the prediction accuracy of the relevance vector machine are studied.The bat algorithm is used to select the appropriate kernel parameters leading to the auto-adaptation of nuclear parameters.The BA-RVM algorithm is verified under the data set used in this paper.The normalization method for the data set in this paper is selected by the comparative analysis of multiple experiments.Secondly,having empirically selected the appropriate kernel function,the BA-FRVM prediction algorithm is proposed by optimizing the training efficiency of the BA-RVM.The effectiveness and reliability of the proposed algorithm are verified by comparing the experimental results using three different types of real data from the UCI website.The training time,the result model vector number and the error rate under the typical fault data of the car are provided in the experiments to compare the performance of the BA-RVM and the proposed algorithm.We analyzed the relationship between the number of iterations and the error rate,the influence of e number of different bats on the prediction accuracy,the relationship between the error rate of the prediction result and the width of the kernel parameter,influence of different training samples on predictive accuracy.In addition,we also performed the comparative analysis of BA-SVR,BA-BP,and the proposed algorithm according to training time,model vector number and error rate.Finally,the BA-FRVM algorithm is implemented by using the Java language and the Java third party libraries matrix jblas,which is applied to the actual number of faults pieces prediction system.To summarize,the experimental results demonstrate that BA-FRVM algorithm has the advantages of less training time,high prediction accuracy,and can better predict the number of cars for typical fault parts by a comparison with BA-SVR and BA-BP algorithm.