Research On Conflict Resolution Method For Industrial Big Data

With the progress of information technology,huge amounts of data have accumulated in the industrial field.The analysis of data to obtain the value contained in the data requires high data quality.An important way to improve data quality is to resolve conflicts.The purpose of this paper for conflict resolution is to improve the data quality of big data in the manufacturing industry and do the necessary preliminary work for subsequent analysis.Regarding conflicts,it is the inconsistent expression description of the same entity.The method of this paper is considered from the perspective of high efficiency and data size.The conflict resolution in big data of product manufacturing(the following industrial big data is big data under this industry)is divided into three aspects.In the first aspect,the data contains not only the exact value(among others,the fuzzy value of the expert's experience and description of the machine's failure level),but also corresponds to the previous market research and after-sales links in the enterprise.To solve this problem,we use fuzzy partial order decision and conflict resolution to find the true value.This step combines the characteristics of the decision-making of fuzzy partial ordering and the characteristics of redistribution and dissolution,and presents them in the form of a relational two-dimensional table.This method takes the fuzzy data in industrial big data into account and uses certain rules to convert it into calculations with precise values,which makes the results more compelling.Secondly,the exact data of the daily internal operations of machine data of industrial big data,due to the high stability of current industrial production,these data have a low value density,that is,most of the data is smooth except for failures,showing the entire production chain.The normal state of business operations.And the data in the industrial field is not so different from the Internet data.In response to this,it is also considered that data growth is fast and different data sources have different authorities for different data to construct a dynamic voting method.This method is a combination of a random forest classifier and a voting algorithm that takes into account data source authority and support between descriptions to both ensure speed and ensure the accuracy required in industrial big data.The third aspect,that is,the study of the parallelization of traditional methods that are currently under consideration,aims to enhance the execution ability of the method and algorithm for handling problems.To solve the key issues of big data research,this paper uses hadoop as a parallelization platform to combine dynamic voting algorithms with parallelization to further improve the operational efficiency.