| 摘要 |
PURPOSE:To efficiently discriminate and predict (automatic discrimination) of a new sample by utilizing a canonical correlation analysis method and converting classification data to weighing value characteristic data and then selecting an optimum explanation variable using regression analysis. CONSTITUTION:Data consisting of a plurality of factors and a group of data divided into a plurality of groups are input 10 for constituting as a canonical correlation analysis model and classification data are converter 12 to characteristic value (weighing value) data using the canonical correlation method. Then, a regression analysis utilization means 120 first associates the characteristic value data with a variable to be explained of the calibration analysis model, associates them with the explanation variable of data consisting of a plurality of factors obtained by the input process 10, and selects 13 the explanation variable which is effective for the characteristic value according to a specific standard. Then, an analysis model expression is created 18 using a calibration expression consisting of a selected explanation variable. Use of the analysis model expression achieves a prediction 19 for discriminating to which group data belong. |
