A Feasible and Adaptive Water-Usage Prediction and Allocation Based on a Machine Learning Method

The definition of entitlements over shared water resources is not an easy task and requires the use of specific decision tools in order to reach the maximum level of objectivity. To develop such tools, the knowledge of the effective water-use by different consumer groups is required. In this paper, such a tool will be introduced. It is based on an adaptive concept which uses previous effective water requirement (EWR), previous water allocated, environmental and socio-economic conditions to derive policies to determine water to be allocated in the future. The model combines two water demand modeling approaches, a micro-component-based approach to calculate the EWR and a data-based approach to calculate and predict the future water-use if no policyadaptations are made. The results of the two models are used subjected to environmental and socioeconomic conditions to determine future policies of water allocation by induction. Due to the uncertainty of the determinants of water demand, fuzzy logic theory is applied. This allows imitating the behavior of water resources administrators. The model can be a decision aid for water distribution planning and such an instrument can be very useful for the local specialists and administrations for justification of their allocation practice.