International Journal on Advanced Computer Theory and Engineering

Abstract

Automated code generation using machine learning techniques has emerged as a transformative approach to software development, enabling developers to generate high-quality code with minimal manual effort. This paper explores recent advancements in machine learning-driven code generation, focusing on deep learning models, transformer-based architectures, and large language models (LLMs) such as GPT, CodeBERT, and Codex. Key methodologies include natural language processing (NLP) for translating human-readable descriptions into executable code, reinforcement learning for improving code efficiency, and fine-tuning techniques to enhance model adaptability. We discuss various applications, including code completion, bug fixing, and optimization, while addressing challenges such as code correctness, security vulnerabilities, and domain-specific adaptations. Finally, we highlight future directions, emphasizing the need for explainability, improved dataset quality, and hybrid AI-human collaboration for robust and efficient automated code generation.