A Multi-Source Dataset for CS1 Failure Prediction in a Sub-Saharan African Context

A comprehensive, anonymised dataset collected from 284 first‑year students enrolled in a CS1 course (Computer Systems and Architecture) across four academic cohorts (2017/18–2020/21) at the University of Zambia, a large public university in sub‑Saharan Africa. The dataset integrates four sources: (1) student information system records (demographics, sponsorship, programme, minor, course workload); (2) a pre‑course survey (self‑reported computing experience, prior training, motivation, computer ownership); (3) Moodle learning management system logs (17,870 raw events aggregated to daily unique hits and component‑specific counts); and (4) assessment records (20 weekly quizzes, four semester tests, and a final examination). Derived features include a weighted academic momentum score, binary flags for LMS activity and self‑sponsorship, a COVID‑19 cohort indicator, and numerical encodings of survey responses. All personally identifiable information has been anonymised through hashing and pseudonymisation. The dataset supports a wide range of learning analytics applications, including early warning system development, replication of predictive modelling, cross‑institutional comparison, engagement pattern analysis, and meta‑analysis of CS1 failure factors. Data are provided under a CC‑BY‑NC‑ND licence. The dataset is publicly available in the Mendeley Data repository.