A novel double incremental learning algorithm for time series prediction

• Published in: Neural computing & applications Vol. 31; no. 10; pp. 6055 - 6077
• Main Authors: Li, Jinhua, Dai, Qun, Ye, Rui
• Format: Journal Article
• Language: English
• Published: London Springer London 01.10.2019; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Science; Data Mining and Knowledge Discovery; Image Processing and Computer Vision; Iterative methods; Machine learning; Original Article; Performance prediction; Probability and Statistics in Computer Science; Support vector machines; Time series; Traveling salesman problem; Incremental learning; Time series prediction (TSP); Incremental SVM; Double incremental learning (DIL) algorithm
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-018-3434-0

Based on support vector machine (SVM), incremental SVM was proposed, which has a strong ability to deal with various classification and regression problems. Incremental SVM and incremental learning paradigm are good at handling streaming data, and consequently, they are well suited for solving time series prediction (TSP) problems. In this paper, incremental learning paradigm is combined with incremental SVM, establishing a novel algorithm for TSP, which is the reason why the proposed algorithm is termed double incremental learning (DIL) algorithm. In DIL algorithm, incremental SVM is utilized as the base learner, while incremental learning is implemented by combining the existing base models with the ones generated on the new data. A novel weight update rule is proposed in DIL algorithm, being used to update the weights of the samples in each iteration. Furthermore, a classical method of integrating base models is employed in DIL. Benefited from the advantages of both incremental SVM and incremental learning, the DIL algorithm achieves desirable prediction effect for TSP. Experimental results on six benchmark TSP datasets verify that DIL possesses preferable predictive performance compared with other existing excellent algorithms.