PURPOSE: To perform inference about seasonal variation while keeping the inference precision by performing inference by generated classification models and changing the classification reference of instance data in accordance with the precision of inference results.

CONSTITUTION: A precision evaluating part 8 compares inference values based on classification models of respective seasons with preliminarily set values to evaluate the precision of inference values; and if the precision is judged to be insufficient, a data classifying part 2 is instructed to set one month as a unit as the classification reference of data for modeling. Thus, the data classifying part 2 classifies data for modeling as data of each month, and a model generating part 5 generates a classification model for each month. An inference device 7 uses data for verification to calculate the inference value based on the classification model for each month, and the precision evaluating part 8 compares the precision of this inference value based on the classification model of each month with a preliminarily determined set value; and if the precision is judged to be sufficient, this classification model of each month is stored in a final model storage part 9.