Data Mining on Forecast Raw Water Quality from Online Monitoring Station Based on Decision-Making Tree

The excessive propagation of algae caused by eutrophication of aquatic environment in the urban source water supply is the main issues of concern to drinking water purification industry in recent years. The prediction of algae in raw water can offer time guarantee for operation of contingency caused by excessive propagation of algae which can ensure the safety of water supply. In the study, we collected 115 daily measured data about indirect monitoring of raw water quality of algae and solar irradiance data from online monitoring and direct-line artificial monitoring of chlorophyll content of raw water. We select decision-making tree which is very visible and easy realized as data mining tools, and set up decision-making tree model which is used to predict the level of chlorophyll in raw water in next day. To enable online monitoring data and artificial monitoring data with the same dimension, combined with the algal growth dynamics, we transform several on-line monitoring data of dissolved oxygen and solar irradiance data in one day into one data per day, that is mean calculating the average standard deviation and average. The former 100 sets of data are used to train and set up decision-making tree model which is to predict the level of chlorophyll in next day. The rest 15 sets of data are used to test data. The results of simulation show that the prediction accuracy can reach 80%.