Calibration And Evaluation Of Government Procurement E-commerce Big Data Based On Machine Learning

The government procurement mall is a procurement platform for governments on the Internet,and it processes a large amount of supply e-commerce data every day.Faced with the problems of different data sources,numerous categories,and no unified writing format,the traditional processing methods are not only time-consuming and laborious,but also the processing results are not unsatisfactory.Based on machine learning,this paper conducts research on the acquisition,calibration,and evaluation of government procurement e-commerce big data to achieve rapid acquisition of political and mining data,accurate calibration of the same product,and effective prediction and evaluation of new supplier quotes using historical prices of the same product,which can promote the application of machine learning in the field of government electronic procurement,assist the government to intelligently monitor the quality and price of commodity,reduce human factor interference and management costs,reduce the price of procurement transactions,improve procurement efficiency,enhance procurement timeliness,and ensure win-win cooperation among the government procurement mall,supplier e-commerce and purchasers.Therefore,the main work of this paper is as follows:Firstly,after analyzing the diverse sources and differentiated characteristics of government-purchased e-commerce big data,the data collection program designed and implemented is used to directionally and quickly obtain e-commerce data on the web pages of various government procurement malls.In the process of program acquisition,it can avoid the differences between the pages of the government procurement mall,can automatically filter duplicate web pages,can automatically sort out all kinds of commodities,and store the obtained e-commerce data by category name,which is convenient to query and recall the saved data through many different forms.The experimental results show that more than 200,000 effective data of government procurement e-commerce can be collected and updated in real time every day,providing data support for subsequent commodity identity calibration,price prediction and rational evaluation.Secondly,a commodity identity calibration model based on Long Short Term Memory Network(LSTM)is proposed.The model is composed of three sub-models such as word segmentation,importance ranking,and similarity calculation.The word segmentation sub-model preprocesses the e-commerce big data to obtain a differentiated keyword sequence,the LSTM importance ranking sub-model screens the most important keyword sequences that characterize the product information,The LSTM similarity calculation sub-model accurately calibrates the same commodity in the given big data.In addition,binary search,Glo Ve word vectorization,and word sequence semantic verification technology are introduced to improve the calibration speed,training sample utilization rate,and high calibration generalization ability,respectively.The experimental results show that,when dealing with big data of different types of government procurement e-commerce,the accuracy of calibrating the identity of confusing samples is high.Finally,the prices calibrated as the same commodity are visually preprocessed,and clustered in three ways.The price clustering results show that DBSCAN is more suitable for removing price singularities than K-means and hierarchical clustering.Based on the extended Dickey-Fowler(ADF)test,the quotation is divided into constant terms,stationary and non-stationary time series.For each series,autoregressive integrated moving average(ARIMA)model,support vector machine(SVM),Gaussian process(GP)model and Gaussian process mixture(GPM)model are used to predict the quotation and evaluate the rationality.The experimental results show that the GP model and the GPM model can output confidence intervals for quote prediction,making the prediction and evaluation results more credible.In addition,the prediction accuracy of the GPM model is generally higher than the other three models,and it is an effective model for commodity price prediction and evaluation.