Research On Track Data Mining For Production Record

With the increasing popularity of location acquisition technologies such as GPS,RFID,barcodes,and two-dimensional codes,it is possible to collect trajectory data of moving objects,and trajectory data mining has become a major branch in the field of data mining.At present,most of the trajectory mining research at home and abroad is concentrated in the field of geographic environment,and there is little research in the field of manufacturing.The workshop production data that supports RFID contains a lot of rich information as special trajectory data,but the data quality is poor,the correlation between the data is complicated,and the lack of corresponding means to deal with these data has not been effectively used.Cycle forecasting is one of the most critical issues in production planning and is also often involved in trajectory mining,but production cycle time prediction methods are more limited to statistics and simulation.Therefore,the research content of this article is divided into two parts.First,a data cleaning model based on production data was established to improve data quality and solve data quality problems.Second,from the perspective of practical application,based on the radial basis function network model,combined with the production data is the characteristics of trajectory data to predict the production cycle of products.The details are as follows:Aiming at the problems of data anomalies and redundancy in the quality of production data,firstly,clear definitions and cleaning rules are given for the researched production data quality problems,data cleaning content and cleaning technology.Then a cleaning model for production data is proposed.The model is based on anomaly and redundancy detection,and the idea of clustering is introduced to achieve similar records in the vicinity and achieve rapid detection.Aiming at the abnormal and redundant data detection links in the model,a data quality detection and correction algorithm is proposed.The actual data is used to compare the algorithm with the traditional detection algorithm to verify the efficiency and effectiveness of the algorithm.Aiming at the problem of low production cycle prediction efficiency of products,firstly,the main factors of the production logistics system are analyzed,and then the key factors related to the cycle time are searched to construct candidate feature sets;Define logistics trajectory from raw data,quantitatively evaluate WIP inventory,machine utilization,product production time,etc.,and convert unusable production operation records into numerical data such as product inventory,machine utilization,product production time,etc.to fill Candidate feature set.Secondly,the radial basis function neural network is optimized according to the characteristics of production data,that is,the parameters of the network are estimated by information entropy clustering.Finally,this paper uses two experiments to prove the effectiveness of the optimization method.First,compare the average absolute error and root mean square error of the optimized method with the original method on the standard data set.Second,use the actual data set to compare the average absolute error and root mean square error of different methods.The experimental results show that the optimized method can predict the results more accurately.