摘要 |
<p>PROBLEM TO BE SOLVED: To forecast large/small fluctuations of time series data in a transaction market with strong nonlinearity by small parameters. SOLUTION: This time series data forecast device is provided with an arithmetic part 4 controlling to calculate a transition probability of a state X(t) in a present time t, a state X(t -1) in a time t-1 a unit time before the present time t, a state X(t-2) in a time t-2 two unit times before the present time t, and a state X(t-3) in the time t-3 three unit times before the present time t based on respective parameters p, q, and w stored in a p storage part 1, a q storage part 2, and a w storage part 3 by referring to a memory 5, forecast a state X(t+1) in a time t+1 a unit time after the present time t from the transition probability, and output the forecasted state X(t+1).</p> |