Loess Compressibility and Soil Property Regression Models Summary

A 5.5 KB dataset summarizing regression models for loess compressibility and soil properties. It includes results from models like Random Forest, MLP, CART, and SVM, with analysis of error metrics and correlations. The study provides data from Huocheng and Xinyuan counties in Xinjiang.

Use Cases

• Compare performance of Random Forest, MLP, CART, and SVM models on predicting loess compressibility parameters.
• Analyze correlations, such as 'significantly positively correlated' or 'negatively correlated', between soil mechanical parameters and compressibility.
• Evaluate model error metrics priority for engineering classification of loess soils in the study areas.

Strengths

• Includes results from multiple machine learning models (Random Forest, MLP, CART, SVM).
• Focuses on specific geographic study areas: Huocheng and Xinyuan counties in Xinjiang.
• Dataset is openly licensed under CC BY 4.0.

Limitations

• Extremely small dataset at only 5.5 KB, indicating a summary rather than raw data.
• No column names or sample data are available, limiting direct analytical use.
• The scope is limited to a summary of model results, not the underlying soil measurements.

Provenance

Source

figshare

Freshness

Last updated March 23, 2026.

Geography

Huocheng and Xinyuan counties in Xinjiang, China.