Dataset for input

This dataset comprises 108 geochemical samples from four outcrop sections (Dahuangshan, Jingjingzigou, Hongyanchi, and Guodikeng) of the Permian Lucaogou Formation in the southeastern Junggar Basin, China, supplemented by 41 samples from three additional sections (XDLK, XLK, YMS) for validation. Measurements include total organic carbon (TOC), major/trace/rare earth elements, and organic biomarkers (e.g., pristane/phytane, dibenzothiophene/dibenzofuran), obtained via Rock‑Eval pyrolysis, X‑ray fluorescence, ICP‑MS, and GC‑MS. The data were collected to investigate lateral heterogeneity in organic matter enrichment and to identify dominant paleoenvironmental controls (paleoredox, paleoclimate, terrigenous input, etc.) using machine learning algorithms, thereby supporting shale oil exploration in lacustrine rift systems.