Prediction Of Cryptocurrency Prices Using LSTM, SVM And Polynomial Regression (2403.03410v1)

Abstract: The rapid development of information technology, especially the Internet, has facilitated users with a quick and easy way to seek information. With these convenience offered by internet services, many individuals who initially invested in gold and precious metals are now shifting into digital investments in form of cryptocurrencies. However, investments in crypto coins are filled with uncertainties and fluctuation in daily basis. This risk posed as significant challenges for coin investors that could result in substantial investment losses. The uncertainty of the value of these crypto coins is a critical issue in the field of coin investment. Forecasting, is one of the methods used to predict the future value of these crypto coins. By utilizing the models of Long Short Term Memory, Support Vector Machine, and Polynomial Regression algorithm for forecasting, a performance comparison is conducted to determine which algorithm model is most suitable for predicting crypto currency prices. The mean square error is employed as a benchmark for the comparison. By applying those three constructed algorithm models, the Support Vector Machine uses a linear kernel to produce the smallest mean square error compared to the Long Short Term Memory and Polynomial Regression algorithm models, with a mean square error value of 0.02. Keywords: Cryptocurrency, Forecasting, Long Short Term Memory, Mean Square Error, Polynomial Regression, Support Vector Machine

• The paper presents a comparative analysis of LSTM, SVM, and Polynomial Regression to minimize prediction error in cryptocurrency prices.
• The study found that the SVM model with a linear kernel achieved the lowest MSE of 0.02, outperforming the other methods.
• Researchers employed an 80:20 train-test split and advanced hyperparameter tuning to validate the robustness of each predictive model.

Comparative Analysis of LSTM, SVM, and Polynomial Regression for Cryptocurrency Price Prediction

Introduction

The paper presents a rigorous examination of various algorithms intended for the prediction of cryptocurrency prices, a subject of burgeoning interest given the volatile nature of digital currencies. The authors, Novan Fauzi Al Giffary and Feri Sulianta, concentrate their efforts on evaluating Long Short Term Memory (LSTM), Support Vector Machine (SVM), and Polynomial Regression algorithms to determine which method provides the most accurate forecasts. Their research emanates from the necessity to devise reliable predictive models amidst the uncertainties that typify cryptocurrency investments.

Methodological Approach

The implementation of the research was structured around a series of well-defined stages, encompassing data collection, preprocessing, allocation, model design, training, testing, and eventual comparison of results through Mean Square Error (MSE) analysis. Utilizing Python as the programming suite of choice, the paper's dataset comprised Bitcoin price information from 2020 to 2024, sourced from the Yahoo Finance portal. Notably, the investigation employed an 80:20 data split ratio for training and testing purposes, respectively, an approach that aligns with standard practices in machine learning research.

Model Exploration and Implementation

LSTM

The LSTM model, a derivative of Recurrent Neural Networks, was recognized for its superior ability to model time series data, making it an apt choice for this investigation. Through meticulous testing across varying epoch counts, the research pinpointed the optimal configuration that minimized MSE values, underscoring the model's potential for accurate price prediction.

SVM

Transitioning to the SVM model, the research delved into both linear and non-linear kernels, exploiting the GridSearchCV method for hyperparameter tuning. This approach facilitated a comprehensive exploration of parameter spaces, ultimately revealing a configuration that achieved remarkably low MSE values, signaling SVM's robustness in forecasting cryptocurrency prices.

Polynomial Regression

Finally, the Polynomial Regression model was scrutinized for its capacity to address non-linear data trends. By adjusting the degree parameter and employing rigorous testing, the paper ascertained the model's efficacy relative to the complexity of the cryptocurrency market's dynamics.

Comparative Analysis and Results

The comparative evaluation of the three algorithms unveiled that the SVM model, employing a linear kernel, outperformed its counterparts by registering the lowest MSE value of 0.02. Conversely, Polynomial Regression exhibited the highest MSE, suggesting a relatively inferior predictive capability within the context of this research. These findings are instrumental in endorsing the SVM model's superiority in cryptocurrency price prediction scenarios.

Conclusion and Future Directions

The paper conclusively determined the Support Vector Machine's predominance in forecasting cryptocurrency prices with minimal error. It suggests an expansive path for future research, including the incorporation of additional variables such as market sentiment analysis to enhance predictive accuracy. This paper not only contributes significantly to the field by guiding researchers towards algorithms with proven efficacy but also opens the door for further explorations aimed at refining cryptocurrency price prediction models.

The implications of this research extend beyond academic inquiries, offering practical insights for investors, financial analysts, and policymakers interested in the intricacies of cryptocurrency markets. Moreover, it sets a foundation for the development of sophisticated tools that could potentially stabilize the inherently erratic nature of digital asset investments.