PURPOSE: To automatically detect a stock price pattern from an unspecific data sequence by using a multi-layered reverse propagation network model part which outputs a data sequence showing a candidate for the stock price pattern corresponding to the stock price data sequence and the stock price data sequence indicating a candidate for corresponding stock price data to an output layer.

CONSTITUTION: The multi-layered reverse propagation network model part is used which outputs the stock price data sequence consisting of ≥1 stock price data to an input layer and outputs the data sequence indicating the candidate for the kind of the stock price pattern corresponding to the stock price data sequence and the stock price data sequence indicating the candidate for the corresponding stock price data to the output layer. Namely, a reverse propagation network model with a tutor after learning is used for the normalized stock price data sequence to recognize the stock price pattern. Therefore, even the stock price data sequence which differs in brand and period, and the features of a vague stock price pattern is learnt by the reverse propagation network model to eliminate the complexity of rule description, etc. Consequently, the vague stock price pattern which is already learnt can automatically be recognized.