Research And Application Of Outlier Detection In Analysis Of Mobile Communication Data

Potentially meaningful knowledge can found from massive data by Data mining techniques, it’s an important challenge to find few abnormal behavior of the object from a number of data and find a meaningful pattern from these behaviors. There are often some data Points called outliers, which are different from normal Points or the Pattern of dataset in the real life. As we know that the normal behavior is much more than abnormal behavior, but a small number of abnormal behaviors may have hidden an interesting knowledge. Therefore, study of outlier has some basis theoretical and practical significance.In this paper, makes a detail study for outlier detection method, combined with cluster analysis and outlier detection technology in data mining, at the same time, makes an analysis for research and application of the outlier detection method at home and abroad, gives the model of outlier detection application. Then introduce the use of data preprocessing, it can improve the overall performance of detection, and finally gives partition-based outlier detection method, applied to the field of wireless network communication data analysis. The main research work includes:1. Making an analysis for the research background and research status of outlier detection method at home and abroad, its range of applications as well as its advantages and disadvantages are analyzed in detail, get the relevant conclusions, outlier detection technology is applied to the mobile communication field of data analysis.2. According to the knowledge and the application of outlier detection, given the basic model of outlier detection, and a detailed analysis of each of its component parts.3. Because of the inconsistent specification of the original data set and large data, making a detailed analysis of the original communication data sets, and through a detailed data-preprocessing of the data set data cleaning, attribute field to select, format conversion and finally obtain high quality to be detected data set, in this way can improve the efficiency of the outlier detection.4. According to the model of outlier detection, we combined with the needs of the outlier detection applications, a method of based on the division outlier detection technology is proposed, make a detailed study of them, and apply it to the mobile communication data analysis.5. Making an experiment test about the proposed method of based on the division outlier detection technology, and making a detailed analysis of the experimental results.In this paper, according to the model of outlier detection, a method of outlier detection based on the division is shown, combing with the cluster analysis and outlier detection technology, applying the pruning methods and fuzzy processing technology, and applied it to mobile wireless network communication data analysis. Through the experimental results of Detection accuracy and the detection time for the method of the performance, shows that the method has good detection results and the overall performance is relatively high.