Replication Package: Empirical Evaluation of Recurrent Neural Networks for Agile Effort Estimation
This repository provides the open-science replication package, experimental benchmarks, and decision-support prototype for the empirical research study "Empirical Evaluation of Recurrent Neural Networks for Agile Effort Estimation: A Realistic Benchmark and Decision-Support Prototype" . The package contains: (i) deterministic partition definitions mapping all 23,313 issues across 16 Deep-SE projects into chronological splits; (ii) per-issue test predictions across naive heuristics, statistical baselines, recurrent models, and factorial runs (N = 4,671); (iii) automated reproduction scripts generating all manuscript tables and non-parametric s
Resource information
- Format
- zip
- Original repository
- 23788447
- DOI
- 10.5281/zenodo.23788447
- Credit basis
- Creators in Zenodo record metadata
- Access
- Open files available from the original Zenodo deposit
- Last checked
- Nov 5, 2026
- License
- mit-license
- Last record update
- Nov 5, 2026
About this resource
This repository provides the open-science replication package, experimental benchmarks, and decision-support prototype for the empirical research study "Empirical Evaluation of Recurrent Neural Networks for Agile Effort Estimation: A Realistic Benchmark and Decision-Support Prototype" . The package contains: (i) deterministic partition definitions mapping all 23,313 issues across 16 Deep-SE projects into chronological splits; (ii) per-issue test predictions across naive heuristics, statistical baselines, recurrent models, and factorial runs (N = 4,671); (iii) automated reproduction scripts generating all manuscript tables and non-parametric statistical tests (Wilcoxon signed-rank and Vargha-Delaney effect sizes) in under ten seconds; (iv) standalone factorial ablation training scripts across independent random seeds; and (v) complete source code for the FastAPI inference microservice and React decision-support prototype.