A CEMS (Continuous Emission Monitoring System) is a package of equipment monitoring the stack emission all day. The central database with a friendly data query interface could be successfully established via receiving monitoring data from the CEMS equipment. However the central database can't support a strategic decision without categorizing the monitoring data, for example we would like to know factors of stack emission predictor. This paper is a research of categorizing the CEMS monitoring data as a data warehouse and surveying data mining system. We design a schema of the data warehouse according to characteristics of the CEMS monitoring data, and then we accomplish data transformations during the process of extracting data from the source data and establish a regional CEMS data warehouse. By using neural network mining techniques and considering the characteristics of the monitoring data, we try to find out useful factors of stack emission predictor. We using the most famous mining tool-the "IBM Intelligent Miner" as our mining tool. The mining result shows whether training or testing predict results that highly correlate with the original data.