PROBLEM TO BE SOLVED: To provide a time sequential data predictive device with which precise prediction can be performed even concerning the future losing a decision theoretic character. SOLUTION: After time sequential data are fetched by a data input part 1, these data are stored through a data embedding part 2 into a 1st attractor storage part 3. The data from the data input part 1 are worked through a data working part 4 and embedded into the data embedding part 2. In this case, embedding is performed while using one piece of data for N pieces of data, and the result is stored in a 2nd attractor storage part 5. The attractors in the 1st and 2nd attractor storage parts 3 and 5 are supplied to a nearby point search part 6, afterwards, the attractor prepared from rough data stored in the 2nd attractor storage part 5 is used until near the desired time point of prediction in the future, and rough prediction is performed by applying this result to a predictive part 7. When the prediction gets closer later, search is performed while using the attractor stored in the 1st attractor storage part 3, and the prediction is executed by applying this searched result to the predictive part 7. The predicted result is outputted from a result output part 8.
展开▼