Abstract : In the software development landscape, code review plays a vital role in maintaining high code quality and ensuring the reliability of software products. This is particularly crucial in the field of data science, characterized by sophisticated algorithms, intricate data pipelines, and a relentless pursuit of model accuracy. However, conventional code review practices often struggle to meet the unique demands posed by data science projects. This paper examines the intricacies of code review within the context of data science, emphasizing its distinct challenges and advocating for more advanced, automated solutions. We present code2Prompt, a novel tool specifically designed to harness the capabilities of Large Language Models (LLMs) for improving code review in data science. Through a rigorous comparative analysis with prominent LLM-based tools like GitHub Copilot, DeepCode, and AI21 Labs' Code Review, we showcase code2Prompt's superior ability to provide contextual code comprehension, prioritize critical code segments for review, and transcend conventional bug detection by offering a holistic assessment of code quality. These features are essential for streamlining and optimizing data science workflows. Our study aims to provide a comprehensive understanding of code review's importance in data science, examining its unique opportunities and challenges. We meticulously illustrate code2Prompt's functionality within the data science code review process, emphasizing its architectural underpinnings and design principles. Furthermore, we employ a robust methodology and a diverse dataset of real-world data science projects to conduct a comparative analysis of code2Prompt against established LLM-powered tools, evaluating their respective strengths and weaknesses across key performance indicators. By analyzing the collected data, we provide valuable insights into the relative merits and limitations of each tool, particularly their effectiveness in addressing the specific challenges inherent to data science code review. We conclude by discussing the broader implications of our findings, highlighting code2Prompt's potential to significantly enhance the efficiency, quality, and overall success of data science projects. We also delineate avenues for future research and development, including integrating code2Prompt within various data science workflows, addressing potential ethical considerations, and advancing the frontiers of LLM-powered solutions for software development.

Keyword : Code Review, Code Quality, Large Language Model, Data Science Workflow.