The Method To Find The Minimum Testing Workload Of Auxiliary Test Unit Based On Data Mining

With the great development of communication industry, the scale of mobile communication network is becoming larger and larger, and the network is increasingly complex with the improvement of2G and3G network and the development of large-scale commercial4G network. In order to accurately grasp the real-time network status and feel network performance from the perspective of the user, the network optimization has become an inevitable choice for mobile operators.ATU (Auxiliary Test Unit) has been an important and indispensable tool for mobile network optimization, but the current test schedule that the operators carried out has no theoretical basis, and is only at a specific time and a particular region, which has brought a series of problems. Such as how long is the minimum testing period and how much is minimum testing workload, which indicators are the most effective indicators for reflecting network status. If testing period is too long or the amount of the sample data is too small, it cannot reflect the network status and causes a decline in user satisfaction; If testing period is too long or the amount of the sample data is too large, it is a waste of human and material resources, increases network maintenance costs and reduces test efficiency. The method to find the minimum testing period of ATU has been completed by other student in our lab, the main achievements to deal with ATU problems in this paper are as follows:1. The analysis of the development of mobile communication and network optimization method, highlight the advantages and problems of ATU, devise research scheme of minimum testing workload of ATU based on data mining.2. This paper proposes a series of data pre-processing methods of ATU data, including the use of Excel for data cleaning, like automatic fill missing values, construct new properties for easier data mining, the use of Min-Max method for data standardization in order to eliminate differences in data dimension, the use of correlation analysis between the KPI indicators by MATLAB to reduce uncorrelated indicators, the use of data sampling method to enrich amount of data, etc.3. This paper proposes the data mining method to find the minimum testing workload of ATU system. For2G,3G network voice service and data service, draw a scatter plot by Excel to observe data distribution and characteristics of each index, employ nonlinear fitting by OriginPro under a certain level of confidence to predict the status of the testing network by the minimum testing workload. Achievements of this paper can greatly reduce resources invested in network optimization for operators and offer a theoretical basis for the future development of rational testing scheme.