Probabilistic Apriori and Episode Mining Techniques for intelligent Management of Water Supply Networks

A software tool, based on Data Mining techniques, which allows to realize early fault diagnosis, during the remote sensing activity of complex water supply networks, is proposed. In order to foresee and analyze fault and malfunction of water plant, association rules and sequential patterns among events, as warnings and actions, should be discovered. Data Mining techniques such as A-Priori and Episode Mining are suitable to accomplish such task. However, the main difficulty in applying such techniques is the correct interpretation of the results. When the events are highly frequent, the above algorithms return relationships between events that are not correlated on the net-physical level and therefore they are not significant. To overcome such problems it is proposed a novel version of Apriori and Episode mining techniques where the significance of the relationships among events is obtained through probabilistic analysis. The proposed algorithm has been tested making the analysis of the three years old historical data acquired by the remote sensing system of a real water supply network.