Optimization of Support Vector Regression Parameters Using Firefly Algorithm

Please enable javascript in your browser.

Ghanbari, Mohammad Reza | 2020

1550 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 52576 (02)
University: Sharif University of Technology
Department: Mathematical Sciences
Advisor(s): Mahdavi-Amiri, Nezameddin
Abstract:
Support vector regression (SVR) in the field of machine learning attracted much attention because of its attractive features and high efficiency for high-dimensional and nonlinear data. Although support vector regression has shown to be very effective for prediction problems, it is necessary to adjust the parameters contained therein to obtain the desired output with error rates. In the past, this was done manually, by trial and error. Over time and by development of optimization algorithms, one of the newest methods to solve such problems is the meta-heuristic optimization algorithms. Therefore, in this thesis, we use the firefly optimization algorithm, which is a population-based algorithm, to adjust the mentioned parameters. Using the three time series data sets from the daily closed stock price of the three major technology companies, we train and test the SVR model and compare the results with eleven other meta-heuristic algorithms
Keywords:
Support Vector Regression ; Meta Heuristic Algorithm ; Time Series Prediction ; Parameter Selection ; Firefly Optimization Algorithm