Transparent Peer Review By Scholar9
A Review and Taxonomy on Forecasting Crypto Prices Based on Machine Learning and Deep Learning Models
Abstract
Crypto price prediction is a category of time series prediction which extremely challenging due to the dependence of crypto prices on several financial, socio-economic and political parameters etc. Moreover, small inaccuracies in crypto price predictions may result in huge losses to firms which use crypto price prediction results for financial analysis and investments. Conventional statistical methods render substantially lesser accuracy compared to new age machine learning techniques. This machine learning based techniques are being used widely for crypto price prediction due to relatively higher accuracy compared to conventional statistical techniques. This paper presents a review on contemporary data driven approaches for crypto currency forecasting highlighting the salient attributes. Moreover, the identified non-trivial research gap in the existing approaches has been used as an underpinning for subsequent direction of research in the domain. The paper culminates with the performance metrics and concluding remarks.
Archit Joshi Reviewer
04 Oct 2024 02:11 PM
Approved
Relevance and Originality
The article addresses a highly relevant and timely issue in the finance sector—crypto price prediction. As cryptocurrencies continue to gain popularity, understanding their price dynamics is crucial for investors and analysts. The originality is evident in the focus on machine learning techniques, contrasting them with conventional statistical methods. To enhance the originality, the authors could delve deeper into specific machine learning models that have shown promise in this domain, potentially offering unique insights into their applications.
Methodology
The paper outlines a review of contemporary approaches but lacks a clear methodological framework. Providing details on the criteria for selecting studies and techniques reviewed would enhance the credibility of the findings. Additionally, discussing the data sources used for evaluating the performance of different predictive models would offer greater transparency. A systematic review methodology, such as PRISMA, could lend structure to the literature review process and improve reproducibility.
Validity & Reliability
The paper discusses the advantages of machine learning techniques but does not provide specific metrics or data to support claims regarding their accuracy. Including comparative analysis with empirical results—such as accuracy rates, precision, recall, and F1 scores for various models—would strengthen the validity of the findings. Furthermore, addressing potential biases in data selection or model evaluation would enhance the reliability of the conclusions drawn.
Clarity and Structure
The article generally presents its ideas clearly, but the structure could be improved for better readability. Organizing the content into distinct sections—such as "Introduction," "Literature Review," "Comparative Analysis of Models," and "Conclusion"—would enhance the flow of information. Clear headings and subheadings would guide the reader through the various components of the study, making it easier to follow the authors' arguments.
Result Analysis
While the paper mentions performance metrics, it lacks a comprehensive result analysis. Detailed comparisons of the performance of various machine learning models on specific datasets would provide a clearer picture of their effectiveness. Discussing the implications of these results for investors and financial analysts, along with recommendations for future research directions, would enrich the conclusion. This could also include suggestions on how to address the identified research gaps, thus providing a roadmap for future studies in crypto price prediction.
IJ Publication Publisher
Done Sir
Archit Joshi Reviewer