摘要 |
PROBLEM TO BE SOLVED: To provide a feature similarity determining device for evaluating whether or not a sample of time series data as a determination object is similar to a population, effectively determining the similarity to the population for an event difficult to determine a behavior showing the characteristic, and especially determining the similarity based upon the distance relationship between the population and a determination object in a Mahalanobis space of the principal component vector to improve the accuracy of determination. SOLUTION: A determination object sample data is additionally subjected to singular value decomposition by an incremental singular value decomposition part 13, the dimension is compressed by a determination object dimension compressing part 14 to obtain the principal component vector of the determination object. A similarity determination part 16 converts the above vector to Mahalanobis distance to the population, and compares the same with the similarity determination reference to determine whether or not it is similar to the population. COPYRIGHT: (C)2006,JPO&NCIPI
|