Reproducibility Package for Digital Library Recommendation Logs and Analysis

3.6 MB of reproducibility materials for a study on digital library recommendations. The package includes code, data bundles, and outputs for verifying results from the Related-Article Recommendation Dataset (RARD). Author Jian Wang published this package on figshare in April 2026 under a CC-BY-4.0 license.

Use Cases

• Reproducing weighted logit and gradient-boosted tree models based on the described analytical workflow.
• Conducting SHAP-based interpretation of recommendation models based on the provided code and data.
• Verifying robustness analyses for exposure-aware recommendation frameworks as outlined in the manuscript.
• Inspecting sample construction and variable definitions for the Random-20k and Case-control-20k analytical subsets.

Strengths

• Includes three main reproducibility bundles (code, data, outputs) totaling 3.6 MB for full workflow verification.
• Provides integrity files (SHA256SUMS.txt, MANIFEST.json) to ensure downloaded materials are complete and unchanged.
• Explicitly lists Python 3.x dependencies in a requirements.txt file for environment setup.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The data bundle contains derived analytical subsets, not the original full dataset, requiring consultation of the original RARD source.

Provenance

Source

Derived from the Related-Article Recommendation Dataset (RARD).

Collection Method

Archived recommendation-log data, processed into analytical subsets (Random-20k, Case-control-20k).

Time Range

null

Freshness

Last updated 2026-04-30 06:06:16; freshness should be verified.

Geography

null