Research On Product Pricing Optimization Method With Missing Data Part

As people's quality of life improves,so does the demand of customers.In order to better meet the individual needs of customers,manufacturers have to face more small-volume,multi-variety orders,which requires enterprises to quickly and accurately dynamic pricing of products.However,in the face of new products generated by the customer's individualized demand or due to human negligence,data loss often occurs in the process of product dynamic pricing,which leads to the company's inability to reasonably and accurately price the product,which ultimately leads to the company's profit.to reach maximum.In the past traditional dynamic pricing,most of the methods used in the face of missing data are to remove the samples with missing data or to use data interpolation.Although these methods can also obtain prediction results,the performance of these methods is limited and directly removed.Missing data can lead to information loss,especially if the missing data is not randomly distributed,and the bias will lead to errors in the final analysis.Therefore,in order to allow enterprises to dynamically price products in the face of diversified demand,small batches and multiple batches of orders,research on a new product pricing that does not directly process the original data and directly use the missing data samples.The method is urgent.Therefore,based on the analysis of predecessors' research on product dynamic pricing and data missing processing methods,this paper proposes a least squares support vector machine regression algorithm based on gradient information to dynamically price products and directly use missing data.Features are modeled for prediction rather than deleting or imputing data.The method also uses the leave-one-out cross-validation strategy to simultaneously determine the impact of features with missing values on regression accuracy.First,through the reading and analysis of the literature,the factors affecting the price of the product and the processing methods when the data is missing are summarized.Secondly,the related theory and model of support vector machine with good ability to solve small sample size problems and the extended least squares support vector machine(LSSVM)related theory and model are introduced.The error variables are introduced on the basis of LSSVM.It expresses the influence on the regression result when the data is missing.The product dynamic pricing model based on the gradient information-based least squares support vector machine regression algorithm(GELSSVR)is constructed and compared with the LSSVM model algorithm and the complete data prediction.The results were compared to verify the effectiveness of the proposed method.At the same time,the leave-one method cross-validation was used to determine the upper bound of the error variable and the effect of the missing feature on the final regression result.Finally,taking the actual data of a certain enterprise as an example,combined with the proposed product dynamic pricing model to verify the feasibility of the product dynamic pricing model based on the least squares support vector machine regression algorithm of gradient information,for the manufacturing enterprises in the case of data loss Dynamic pricing of products provides support for scientific methods and provides references for other applications that deal with missing data.