A SVM-based Locating Method Of GSM System Uplink Interference

With the advancement of technology, the communication network technology is rapidly developing. Currently, GSM(Global System for Mobile Communication) system is the highest frequency system in the three communications operators, it is necessary to do the maintenance of the system. Situation can’t occur that the poor quality of communication networks leads to the loss of customers. In the GSM system, uplink interference is the main cause call quality and low success rate, it will directly lead to customer complaints, loss and so on. The traditional method of determining the type of uplink interference is that network optimization engineer or experienced personnel analysis the traffic data of cells or detect by the professional measuring instruments. This method is time-consuming, costly and inefficient. Therefore, how to quickly and accurately locate the interference with the scientific method is the current problem to be solved.In according to an idea of theoretical analysis, data preprocessing, feature extraction, algorithm formation, I study the principle of varieties of uplink interference, combined with FAS(Frequency Allocation Support) statistics, and then propose uplink interference intelligent location based on support vector machine. The subject in this paper comes from technology projects of Guangdong mobile company. The technology project is wireless network optimization expert system algorithm base on data mining. The subject has practical significance.The main work of this paper is:(1)Depending on the cause interference, we extracted the 95-dimensional features in FAS data interference.(2)Preprocessing the FAS raw data files, including replacing interference outliers, smoothing interference equality.(3)Because the feature dimension of FAS date is high, and it also has a strong correlation, the data on the FAS, we investigated the characteristics of various kinds of interference and use appropriate statistic to express it. Then we can achieve the purpose of reducing the dimension of the original data. It not only can be summed up well the original variable information, but also can eliminate the interference of irrelevant information system.(4)We study the hybrid model which combine feature extraction and support vector machines, and apply it in disturbance classification. Experiments show that the model has better classification results.The proposed method has been applied a network optimization project in Guangzhou. Practice shows that this method can well determine whether the cell has uplink interference. Compared with traditional methods, this method for determining the efficiency and cost advantages are obvious. In the current information technology, this method can be easily integrated into the software platform, which realizes the process-oriented and intelligent of interference judgment.