Combinations of the multi-model have been successfully used to enhance prediction performance, while fuzzy classification techniques have been claimed to be very useful when interpretability of the generated model is very important. In this paper the combination of the multi-model approach with fuzzy classification methods is proposed for the prediction of domestic water leakage, based on consumer data. Selected fuzzy classification techniques were used to induce classification models to predict cases of possible water leakage originally classified as 'Likely' or "Unlikely'. Based on analysis conducted using the multi-model approach, the 'Likely' cases were reclassified into new classes defined as 'Less likely" and 'More likely'. Comparison of classification outcomes between the original two-class dataset with the modified three-class dataset show very consistent prediction outcomes. The findings suggest the potential of the combined approach for the prediction of domestic water leakage, based on consumer data.
展开▼
