A multidimensional space of variables is divided so that a plurality of highly accurate prediction models can be used according to driving conditions. A cluster division evaluation device according to the present invention obtains an actual measurement value of a value indicating a property of raw water acquired by a water purification plant and an actual measurement value of a value indicating an amount of a drug injected into the raw water. And a clustering unit that divides the points indicating the measured values of the values indicating the properties of the raw water and the measured values indicating the amounts of the drugs injected into the raw water into a plurality of clusters in a multidimensional space, and the measured values. The parameters of the prediction model using the value indicating the property of raw water as the explanatory variable and the value indicating the amount of the drug as the objective variable are optimized for each of a plurality of clusters, and the prediction model with the optimized parameters is output. A regression analysis unit that evaluates the difference between the predicted value of the value indicating the amount of the drug and the measured value of the value indicating the amount of the drug for each number of clusters, and determines the number of clusters based on the evaluation result. It is characterized by having. [Selection diagram] FIG.
展开▼
