Skip to main content
Fig. 3 | Applied Network Science

Fig. 3

From: Using correlated stochastic differential equations to forecast cryptocurrency rates and social media activities

Fig. 3

a Shows the accuracy in predicting whether the time series will increase or decrease for a training period of 120 time steps before the prediction. This is compared to a prediction without using correlated time series. b Here, we show the MAPE (ε) for predicting the value of the time series one time step into the future for 90 and 120 time step training periods. After a small decrease in the MAPE value, there is not a significant change in MAPE as the number of time series increase. c shows the MAPE for predicting a variety of time steps into the future using correlations of 0.4, 0.6, 0.8. d shows the change in MAPE value (Δε) when a time series is added with improved correlation is added. This is done by setting the correlation of the added time series and the predicted time series to the square root of the original correlation

Back to article page