A Comparative Approach between Batch and Online Machine Learning for Predicting Next-Minute Cryptocurrency Price Direction

Barry, Tiernan (2021) A Comparative Approach between Batch and Online Machine Learning for Predicting Next-Minute Cryptocurrency Price Direction. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (2MB) | Preview
Preview	PDF (Configuration manual) Download (2MB) | Preview

Abstract

As machine learning evolves within the era of big-data, this research addressed the topic of comparing batch machine learning against equivalent, online learning techniques in the context of next-minute cryptocurrency predictions. This research designed 3 comparative experiments, each predicting Ethereum, DogeCoin and BinanceCoin’s next minute price direction: (Exp1.a) Batch Logistic Regression vs Online Softmax Regression, (Exp1.b) Batch Decision Tree vs Online Hoeffding Tree, and (Exp1.c) Batch Random Forest vs an Online Adaptive Random Forest. Exp1.a showed that across all 3 alt-coins, the Online Softmax Regression model outperformed the Batch Logistic Regression model, with the best online model accuracy and F1 being 0.61 compared to 0.56 for the Batch Logistic Regression. For Exp1.b, the results marginally favoured Batch Decision Trees over Hoeffding trees in 2 out of 3 datasets, with the best batch accuracy and F1 of 0.56 and 0.44, despite Hoeffding trees performing better on 1 alt-coin (BinanceCoin) with 0.63 accuracy. Exp1.c results favoured the Online Adaptive Random Forest in 2 out of 3 alt-coins, with the highest accuracy and F1 of 0.66 and 0.64 compared to 0.56 and 0.46, while Exp1.c results for DogeCoin marginally favoured the Batch Random Forest. Overall, model performances compare well with existing work on cryptocurrency predictions. Despite mixed results, this paper concluded that there are advantages of deploying online machine learning as opposed to batch learning for predicting next-minute price predictions of Ethereum, Dogecoin and BinanceCoin.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science H Social Sciences > HG Finance > Money > Currency
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Clara Chan
Date Deposited:	11 Nov 2021 16:18
Last Modified:	11 Nov 2021 16:18
URI:	https://norma.ncirl.ie/id/eprint/5134

Actions (login required)

View Item