Data and code for: Source-Traceable Machine Learning of Isothermal Crystallization Kinetics Across Polyester Systems

This deposit contains the frozen dataset, row-level source ledger, analysis code, and full environment documentation accompanying the manuscript "Source-Traceable Machine Learning of Isothermal Crystallization Kinetics Across Polyester Systems," submitted to Polymer Engineering and Science. The dataset comprises 2,873 verified crystallization observations compiled from 198 independent literature studies spanning 47 polyester families. The corrected primary modeling cohort used for the main analysis contains 794 observations from 60 studies and 25 families, filtered to rows with non-missing crystallization temperature (Tc), melting temperature (Tm), and crystallization half-time. A defining feature of this dataset is source traceability: every one of the 2,873 observations is mapped, at the row level, to its originating publication (via DOI or source URL where no DOI exists), the specific table or figure it was extracted from, its verification status, and its data-quality tier. This ledger is included as a standalone file to allow independent auditing of any value in the dataset back to its literature source. Also included is the full analysis notebook covering data consolidation, descriptor derivation, nested leave-one-study-out (LOSO) and leave-one-family-out (LOFO) cross-validation, the primary Ridge regression model, benchmarking against nonlinear algorithms (Random Forest, Extra Trees, Gradient Boosting, XGBoost), random-intercept mixed-effects heterogeneity analysis, and bootstrap uncertainty estimation — along with an environment file documenting exact package versions, random seeds, and hyperparameter grids needed to reproduce every result reported in the manuscript. Files: Polyester_Crystallization_V86.parquet — master 2,873-row dataset Polyester_Crystallization_PrimaryCohort_794.parquet — primary 794-row modeling cohort Source_Ledger_V86_2873rows.csv — row-level source ledger Source_Ledger_StudyLevel_Summary_198studies.csv — study-level summary polymer.ipynb — analysis notebook environment.txt — package versions, seeds, and hyperparameter settings README.md — full file descriptions and manuscript cross-references