Abstract: A compound property prediction apparatus includes a memory storing a training sample library in which a parameter value and a value for a prediction item are preregistered for each individual one of a plurality of training samples and an input device inputting data concerning an unknown sample. Also included is a processor programmed to generate a parameter of the unknown sample, calculate the degree of similarity between the unknown sample and an individual training sample, construct a sub-sample set by extracting training samples whose degree of similarity to the unknown sample is not smaller than a predetermined threshold value, construct a prediction model from the sub-sample set, and calculate the prediction value of the unknown sample based on the prediction model. If necessary, the threshold value is changed until a predetermined minimum number of training samples are included in the sub-sample set.

Abstract: A compound property prediction apparatus includes a training sample library (16) in which a parameter value and a value for a prediction item are preregistered for each individual one of a plurality of training samples; an input device (12) inputting data concerning an unknown sample; a parameter generating device (14) of the unknown sample; a similarity calculation device (18) which calculates the degree of similarity between the unknown sample and the individual one training sample; a sub-sample set construction device (20) which constructs a sub-sample set by extracting training samples whose degree of similarity to the unknown sample is not smaller than a predetermined threshold value; a prediction model construction device (22) which constructs a prediction model from the sub-sample set; and a prediction value calculation device (24) which calculates the prediction value of the unknown sample based on the prediction model.