Anugerah Surya Atmaja, Arie Wahyu Wijayanto

Application of Firefly Algorithm in Determining Hyperparameters on Support Vector Regression to Predict Stock Price with Google Trends

Introduction

Application of firefly algorithm in determining hyperparameters on support vector regression to predict stock price with google trends. Predict stock prices accurately using Support Vector Regression optimized by the Firefly Algorithm, enhanced with Google Trends data. Minimize investment risk with advanced machine learning.

Abstract

Stock fluctuations as well as the tendency to high volatility raise doubts for investors to invest in a company. Efforts that can be made to minimize investment risk are to conduct predictive analysis. The development of machine learning technology and big data can be a support in prediction, one of which is the use of the Support Vector Regression (SVR) method and google trends index data.This research forms a prediction model for PT. BRI (Persero) Tbk. which involves google trend index data using the SVR method. Referring to the constraints in determining the appropriate hyperparameters for the SVR method, the firefly algorithm is used to obtain hyperparameters that optimize the model. Based on modeling, the SVR-FA model involving the google trend index gave the best results, shown by the RMSE and MAPE were 348,47 and 4,12% respectively. This shows that by adding google trend index variables and utilizing machine learning methods in modeling,it will provide better results.

Review

This paper presents an intriguing application of machine learning techniques to address the pertinent challenge of stock price prediction, aiming to mitigate investment risks. The authors propose a model for PT. BRI (Persero) Tbk. that combines Support Vector Regression (SVR) with Google Trends index data, further optimized by employing the Firefly Algorithm (FA) for hyperparameter tuning. The study's focus on enhancing predictive accuracy through advanced algorithmic integration is timely and relevant in the context of fluctuating and highly volatile financial markets. A notable strength of this research lies in its methodological innovativeness. The integration of the Firefly Algorithm to determine optimal SVR hyperparameters is a sophisticated approach to overcome a common limitation of SVR models, potentially leading to more robust and accurate predictions. Furthermore, the inclusion of the Google Trends index data as an external variable is a commendable effort to capture public sentiment or interest, which can significantly influence market dynamics. The reported results, with an RMSE of 348.47 and a MAPE of 4.12%, suggest that the SVR-FA model, augmented by Google Trends, achieves promising predictive performance, demonstrating the value of combining these diverse computational intelligence tools. While the abstract highlights compelling results, a more comprehensive evaluation would benefit from additional details and comparative analyses. To fully appreciate the claimed "best results," it would be valuable to see explicit comparisons against baseline models, such as SVR without FA, SVR without Google Trends data, or other conventional forecasting methods. Future iterations of this work could also delve into the generalizability of this SVR-FA-Google Trends framework across a broader spectrum of stocks and diverse market conditions, along with a more detailed discussion of the dataset's time frame, validation strategies, and the potential economic implications of such a predictive model for investors.

Full Text

Comments

You need to be logged in to post a comment.