| This paper researches into telecommunications data warehouse and data miningtechnology in the field of public security. The author applies cluster analysis andneural network algorithm to mining out special user groups fromtelecommunications data by searching for outliers.First, the author explores into data warehouse architecture and establishes basicdata for the data mining through ETL with CDR (call detail records) as source data.Emphasis is put on the study of neural network algorithm and its test applicationwhile various analysis methods in data mining, such as sorting, clustering,association, and sequential are reviewed.Secondly, the author establishes relevant target users prediction models in theexample of population analysis. Data analysis and mining tool clementine are usedand the improved CRISP-DM data mining process is applied in this process, whichincludes six steps for commercial application data mining: service understanding,data understanding, data preparation, model building, model assessment and modelpublishing. Then these models were put into application to evaluate and test theireffects by searching possible target users.Finally the author briefly discussed the direction for further study.In this paper, variables derived from using CDR, such as calling hours and theproportion of long distance call, are applied in the floating population analysis. Thefloating population prediction among mobile phone users in a certain area hasachieved. satisfactory results.The application of the data mining tool clementine, neural network algorithm,the improved CRISP-DM standard as well as the correctly applied data sampling andderived variables, form the foundation of the whole research in this paper. |
