The Abnormal Data Applied Research In The E-commerce Environment

Anomaly detection as an important part of the data mining application endless dataacquisition, has a wide range of applications in the field of scientific research, stock, financial,insurance, telecommunications and the Internet. The rapid development of the network ande-commerce, making e-commerce enterprises accumulated a large number of customers,enterprises need knowledge and information, outlier is an important part of e-commerceenterprises urgently need to dig out from the large amounts of data, these outlier not payenough attention or as errors, is likely to miss a lot of important information, these abnormaldata can often bring a lot of profit for the enterprise. This paper focuses on key technologiesexpand the e-commerce environment, anomaly detection, and discusses the anomaly detectionin the field of e-commerce applications, and examples for instance background, thee-commerce environment, anomaly detection in customer relationship management. Includethe following:1.Carried out a detailed analysis of the anomaly detection theory. Definition, outliers,anomaly detection and anomaly detection definition and algorithm research situation andsummarized, the advantages and disadvantages of the various algorithms, and finally select adistance-based anomaly detection algorithm As the empirical part of the algorithm.2.Constructed B Internet search company Outlier customers detection index system. Setof15indicators from the last value, the current value and potential value, behavior log data,outreach programs the data reflects the most important customer information, the applicationof factor analysis from the index system filter out the impact of customer composite score sixindicators, six indicators also affect a customer exception of the most important indicators.The reason for the Outlier in the analysis of outlier customers, combined with practicalenough to prove the explanatory power of this study are six indicators.3.The selected Distance-based anomaly detection algorithm and factor analysis. DivisionB50customer index system according to the design, distance-based anomaly detection,02,21,23and29four abnormal customers. Application Pauta criterion on the client factorcomposite score abnormality determination results of distance-based anomaly detectionalgorithm control,21and23two customers are the same, while other customers inconsistentresults.4.For the various Outlier customers screening and analysis of the reason for the Outlier,and to develop appropriate management and marketing strategies. Distance-based anomalydetection algorithm and factor analysis to dig out two exceptions customers, combined withthe analysis of the actual situation, to identify the cause of the exception, to developappropriate coping strategies based on these reasons. For two algorithms inconsistent02,29 customers, the final determination is abnormal client, identified by the distance-basedanomaly detection is stronger. At the same time, the factor composite score higher20,24customers and score a minimum of14,16customers for analysis after analysis shows thatfour customers is not unusual customers, but also gives some management recommendations.