A Multi-Year Machine Learning Analysis of Developer Experience: Influencing Factors, Prediction, and Profile Discovery

Mathew, Alina (2025) A Multi-Year Machine Learning Analysis of Developer Experience: Influencing Factors, Prediction, and Profile Discovery. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (793kB) | Preview

Abstract

It has been established that developer experience (DevX) affects productivity, well-being and tool adoption, but factors that determine its emotional, value-related and perceptual aspects are understudied. Based on the 2022, 2024, and 2025 datasets of the Stack Overflow Annual Developer Survey, this thesis creates an analytical pipeline, end-to-end, to determine the primary determinants of DevX, the predictability and the latent profiles of the developers. The pipeline combines several steps: data harmonisation and MICE imputation; feature selection with the help of FLO that is model-agnostic; joint multi-target prediction with a residual multilayer perceptron; and explainability of the model with the help of SHAP-based feature attributions. Latent Profile Analysis is carried out with Gaussian Mixture Models on calibrated probability vectors and SHAP embeddings as representation space. Bayesian Information Criterion and bootstrap-adjusted Rand Index are used to calculate model selection and profile stability. A similar pattern of behavioural and organisational determinants arises across the three years of the survey with a gradual change to AI-mediated development practices in the recent years. Joint prediction has a high performance (macro-F1: 0.74-0.94), which indicates that DevX is statistically learnable. LPA has a small set of coherent developer profiles that vary in the DevX dimensions, work practices and exposure to the emerging technologies, and are stable in structure over years. These findings show that DevX is empirically structured, predictable and profilable. The results support future DevX-aware interventions, tooling strategies and organisational assessment practices.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Razzaq, Abdul UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Computer software > Computer software - Development T Technology > T Technology (General) > Information Technology > Computer software > Computer software - Development Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	08 Sep 2026 09:22
Last Modified:	08 Sep 2026 09:22
URI:	https://norma.ncirl.ie/id/eprint/9881

Actions (login required)

View Item