Authors:
Sidra Hussain
;
Sheikh Mim
and
Doina Logofatu
Affiliation:
Dept. of Computer Science, Frankfurt University of Applied Sciences, Nibelungenpl. 1, 60318 Frankfurt am Main, Germany
Keyword(s):
Random Forest, Hyperparameter Tuning, Cryptocurrency, Machine Learning, Loss Curves, Decision Tree, Linear Regression, Gradient Boosting.
Abstract:
Cryptocurrency is the current evolving financial market that gives scope for many researchers and machine learning assets to be put into production. Cryptocurrency forecasting is a challenge as following financial assets is difficult due to their volatility and unscalable factor dependencies. This paper puts forward the data of fourteen different types of cryptocurrencies, which will be used to build machine learning (ML) models to forecast the crypto scores in the next fiscal year. For this, four different models are used, and their performance is evaluated and tested. The models are Linear Regression, Decision Tree, Random Forest, and Gradient Boosting. The results obtained from the models show a perfect fit and good hyperparameter tuning, giving evidence of good feature engineering and data scraping. Overall, the models are meant to benefit the financial market immensely by helping to forecast and help investments build in sales and purchasing.