A study on input functions of answer-in-weights neural network for time-series prediction
Haruka Kanada, K. Simizu, Takehiko Ogawa · Society of Instrument and Control Engineers of Japan · 2003
For time-series data prediction, we have investigated a neural network of answer-in-weights structure. For more accurate prediction, we proposed to use an answer-in-weights network consisting of three component estimators. In this paper, we study on input functions of an answer-in-weights neural network. We try to obtain the frequencies of the periodic input functions by frequency analysis of time-series data. To confirm the effectiveness of the proposed method, we make computer simulation using actual economic indicators.