To provide a high-dimensional similarity join method by division capable of efficiently performing similarity join to high-dimensional data without requiring a large amount of storage space in relatively short execution time.

The high-dimensional similarity join method includes a step for dividing a high-dimensional data space and a step for performing join between prescribed data sets, dimensions for dividing the high-dimensional data space and the number of dimensions are predetermined before dividing the high-dimensional data space and each of the cells of the data set are joined only when they are overlapped on the data space or are adjacent to each other.