Optimization is becoming increasingly important to both operators and venders of GSM networks. When the engineers do the optimization, they begin with data from the OMC (Operation and Maintenance Center), and analyze them with their knowledge and experience until they have found the reasons causing the low guideline. We developed an Expert System for this domain. But the amount of the data is very vast, and there exists some information even experienced engineers do not know yet. Moreover, the knowledge of optimization changed soon with the updating of network. Artificially inputting knowledge into the expert system can't work well because of the large burden. So how to make the best of the data and obtain more information from the network became an interesting investigation. In this thesis, we analyzed the current status of optimization for GSM network and briefly described the expert system. Then we defined the purpose of KDD in the optimization as cases. The steps of KDD are defined for the optimization. Fuzzy logic and statistics is used to handle the uncertainty of data. The tasks and verifying criterions of each step are given. The algorithms to find the cases are listed in detail and future work is also given.