A Hybrid System ASVR/NGARCH Tuned by Quantum-Based Minimization to Improve Forecasting Accuracy

Adaptive support vector regression (ASVR) is very useful to act as a predictor for complex time series prediction. However, ASVR cannot avoid volatility clustering and thus worsen its predictive accuracy. Therefore, incorporating NGARCH model into ASVR is schemed for resolving the problem of volatility clustering to best fit the time series. Interestingly, quantum-based minimization algorithm is proposed in this study to tune the resulting weighted-average forecasts between ASVR and NGARCH to improve the forecast performance. Quantum optimization process tackles so-called NP-completeness problem outperforming artificial neural network or quadratic-programming so as to attain optimal or near-optimal result over the scope of search space. It follows that the proposed hybrid system can get the satisfactory results because of highly enhancing its generalization and then improving the accuracy.