Adoption of machine learning (ML) across diverse domains demands a systematic development approach. ML development differs from traditional software development due to its data-driven and iterative nature. Using Kitchenham’s protocol, this systematic literature review (SLR) aims to synthesize existing literature on development processes tailored for ML systems and evolving variants. Models are categorized by lifecycle phases, granularity and support for traceability and continuous deployment. Emerging trends include hybrid and fine-grained approaches bridging academia and industry. Key gaps include limited empirical validation and lack of standardized practices covering ML variants. This review serves as a base to develop reliable and maintainable ML systems.
Jamal et al. (Thu,) studied this question.