| 英文摘要 |
Stock markets occupy a critical position in modern society. The aspiration of every investor is to accurate predict the stock market behavior aiming to maximize his profits. This is a difficult problem because market behavior is volatile, stochastic and affected by many factors such as politics, global economy, investor expectation and others. Deep learning methods have the advantage of learning features automatically through multiple layers of mapping. Support vector machines (SVM) have the advantage of generalizing very well on many different classification and regression datasets. In recent years, deep structured support vector machine networks that successfully combine the superior advantage of both models have received great attention in industry and academia. Comparing with other deep learning networks, deep SVM has many benefits, including (1) the deep SVM has beter regularization ability to avoid overfitting; (2) deep SVM is able to deal with problem of few training samples and high-dimensional feature space. In this paper, we develop a novel deep structured fuzzy dual support vector regression (SVR) machines networks to forecast the market behavior by using the numerical information (stock’s close price) available online. The proposed model is a hybrid model which combines the advantages of: (1) ensemble learning, (2) deep learning, (3) evolutionary optimization, and (4) multiple kernel learning. The proposed method determines the outer and inner bounds of the vagueness region for the estimated result. The proposed deep dual SVR model is able to indicate a level of confidence for the predicted results. The interpretable characteristic for the level of confidence makes the proposed approach more suitable for decision making problem. |
本站僅提供期刊文獻檢索。
