PROBLEM TO BE SOLVED: To divide a multidimensional space of variables so that a plurality of highly accurate prediction models can be used according to the characteristics of data. SOLUTION: The cluster division evaluation device of the present invention has a survey value acquisition unit that acquires a survey value of a trading entity's attribute and a survey value of the trading entity's economic power, and a trading entity's attribute survey value and a trading entity. The point showing the survey value of the economic power of is divided into multiple clusters in the multidimensional space, and using the survey value, the attribute of the trading entity is used as the explanatory parameter and the economic power of the trading entity is used as the objective parameter. The parameters of the prediction model to be used are optimized for each of multiple clusters, and the difference between the forecast value of the trading entity's economic power output by the parameter-optimized prediction model and the survey value of the trading entity's economic power is calculated as a cluster. It is characterized by including a regression analysis unit that evaluates each number of clusters and determines the number of clusters based on the evaluation result. [Selection diagram] FIG. 7
展开▼
