To relatively easily realize product property adjustment and to improve accuracy in property adjustment.

Intermediate property, property control processing conditions and final property in the manufacturing process of a product are stored and held as a data set for every lot in a process management database. As processing 1, the data set for every lot is prepared, and as processing 2, cluster processing is performed on each data set for every lot. As processing 3, using the data set obtained by the processing 2, the casual relation between input and output is quantified by a neural network, considering the intermediate property and property control processing conditions as the input and the final property as the output, to prepare a learning model. In the subsequent model operation, the prepared learning model is used to retrieve the optimum property control processing condition from the intermediate property in the previous process in the property adjusting process.