Stock Price Time Series Prediction using Neuro-Fuzzy with Support Vector Guideline System

Global prediction techniques such as support vector machines show accurate prediction for time series data; however, such models tend to delay the predicted output. Fuzzy systems have benefits in local optimum, thus producing significant results within training sets. Unfortunately, the existing techniques sometimes give undesired effects of surface oscillation at predicted outputs. This paper presents a cascade model called Neuro-Fuzzy with Support Vector guideline system (NFSV) to resolve the problem mentioned above. The proposed model takes benefits from both support vector machine and fuzzy model with appropriate stock price rule filtering. From evaluation, the proposed method seems to have low error rate in stock price time series prediction.