多目標函數仿真機器學習研究－使用基因演算法與遞迴式最小平方估計方法

Machine Learning Approach to Multi-Function Approximation Using GA-RLSE Algorithm

李俊賢(Chunshien Li) ； 陳盈臻(Ying-Chen Chen)

《電子商務學報》 24卷1期 (2022/06) Pp. 1-26

https://doi.org/10.6188/JEB.202206_24(1).0001

摘要

巨量資料使得建立預測模型的時間與計算資源增加，也意謂著建立預測模型的成本提高。因此如何在既運算資源下建立一個高效能的預測模型，是一個重要的工作。本研究提出一種新的方法，以單一複數模糊類神經系統同時預測多個目標，大幅提升模型運作效能。此外本文提出了一個針對多目標的特徵選取策略，可以跨目標篩選出重要的特徵做為系統的輸入變數。在機器學習的部分，本篇研究提出的GA-RLSE混合型學習演算法，以GA優化系統前鑑部參數，以RLSE更新Takagi-Sugeno後鑑部參數。實驗的部分，分別透過同時逼近四個函數與預測兩個時間序列的國外匯率以驗證本研究提出的方法。根據實驗結果與其他方法的性能比較，本研究的方法表現優異。

關鍵字

複數模糊集合 ； 複數模糊類神經系統 ； 基因演算法 ； 遞迴式最小平方估計法 ； 多目標函數仿真

並列摘要

Processing big data to build prediction models requires large computing resources and long periods of time, which increases the cost of building a model. Therefore, developing a high-performance prediction model that requires less computing resources and time is an important task. This paper proposes a novel approach to approximate multiple functions simultaneously in a single complex neuro-fuzzy system. This approach can significantly improve the efficiency of the model. Moreover, the paper presents a feature selection strategy for multiple targets that useful features can be screened out for all the targets. The selected features are then used as inputs to the proposed system. For machine learning, the proposed GA-RLSE algorithm applies in the way that the GA method evolves the premise parameters of the proposed system and the RLSE method updates the Takagi-Sugeno consequence parameters. For experimentation, the proposed approach is tested by simultaneously approximating four functions and predicting two financial time series of foreign exchange rate respectively. With the experimental results and through performance comparison to other methods, the proposed approach has shown excellent performance.

並列關鍵字

Complex fuzzy set (CFS) ； complex neuro-fuzzy system (CNFS) ； genetic algorithm (GA) ； recursive least squares estimator (RLSE) ； multi-function approximation

參考文獻

Anderson, J. A. (1995). An introduction to neural networks. Cambridge, MA: The MIT Press.

Chiu, S. L. (1994). Fuzzy model identification based on cluster estimation. Journal of Intelligent and Fuzzy Systems, 2(3), 267-278.

Holland, J. (1975). Adaptation in natural and artificial systems: An introductory analysis with applications to biology, control, and artificial intelligence. Ann Arbor, MI: University of Michigan Press.

Jang, J. S. R. (1993). ANFIS: Adaptive-network-based fuzzy inference system. IEEE Transactions on Systems, Man, and Cybernetics, 23(3), 665-685.

Kennedy, J., & Eberhart, R. (1995). Particle swarm optimization. Proceedings of the IEEE International Conference on Neural Networks (ICNN’95), Perth, Australia.