The time-series data feature quantity extraction device uses the received unequal-interval time-series data group based on the received input time-series data length and the received observation minimum interval, and whether or not there is a deficiency. The difference between the data processing unit that processes the missing information group that represents and the non-missing element of the matrix of the equidistant time series data group including the missing and the output result element of the output layer of the model for the model And learning a weight vector of each layer of the model, storing the weight vector in a storage unit as a model parameter, accepting time series data of a feature quantity extraction target, and accepting the received feature quantity extraction target By inputting the time series data of the model into the model, the model parameters stored in the storage unit are used to calculate the value of the intermediate layer of the model, And a feature amount extraction unit which outputs the value of the issued intermediate layer as a feature amount representing a temporal change of the data.
展开▼