Enhanced artificial bee colony-least squares support vector machines algorithm for time series prediction

Over the past decades, the Least Squares Support Vector Machines (LSSVM) has been widely utilized in prediction task of various application domains. Nevertheless, existing literature showed that the capability of LSSVM is highly dependent on the value of its hyper-parameters, namely regularization p...

Full description

Saved in:
Bibliographic Details
Main Author: Zuriani, Mustaffa
Format: Thesis
Language:English
English
Published: 2014
Subjects:
Online Access:https://etd.uum.edu.my/4394/1/s93651.pdf
https://etd.uum.edu.my/4394/2/s93651_abstract.pdf
https://etd.uum.edu.my/4394/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Over the past decades, the Least Squares Support Vector Machines (LSSVM) has been widely utilized in prediction task of various application domains. Nevertheless, existing literature showed that the capability of LSSVM is highly dependent on the value of its hyper-parameters, namely regularization parameter and kernel parameter, where this would greatly affect the generalization of LSSVM in prediction task. This study proposed a hybrid algorithm, based on Artificial Bee Colony (ABC) and LSSVM, that consists of three algorithms; ABC-LSSVM, lvABC-LSSVM and cmABC-LSSVM. The lvABC algorithm is introduced to overcome the local optima problem by enriching the searching behaviour using Levy mutation. On the other hand, the cmABC algorithm that incorporates conventional mutation addresses the over- fitting or under-fitting problem. The combination of lvABC and cmABC algorithm, which is later introduced as Enhanced Artificial Bee Colony–Least Squares Support Vector Machine (eABC-LSSVM), is realized in prediction of non renewable natural resources commodity price. Upon the completion of data collection and data pre processing, the eABC-LSSVM algorithm is designed and developed. The predictability of eABC-LSSVM is measured based on five statistical metrics which include Mean Absolute Percentage Error (MAPE), prediction accuracy, symmetric MAPE (sMAPE), Root Mean Square Percentage Error (RMSPE) and Theils’ U. Results showed that the eABC-LSSVM possess lower prediction error rate as compared to eight hybridization models of LSSVM and Evolutionary Computation (EC) algorithms. In addition, the proposed algorithm is compared to single prediction techniques, namely, Support Vector Machines (SVM) and Back Propagation Neural Network (BPNN). In general, the eABC-LSSVM produced more than 90% prediction accuracy. This indicates that the proposed eABC-LSSVM is capable of solving optimization problem, specifically in the prediction task. The eABC-LSSVM is hoped to be useful to investors and commodities traders in planning their investment and projecting their profit.