Lateritic Soil Strength Predictions from 300 Samples in Southwest Nigeria

300 disturbed laterite samples were collected from 30 deposits in Southwestern Nigeria and tested according to ASTM standards. The dataset contains soil index properties used to train and evaluate three interpretable machine learning models for predicting unconfined compressive strength. The work was authored by Lateef Adamolekun and last updated on figshare in May 2026.

Use Cases

• Predicting unconfined compressive strength (UCS) based on easily measurable soil index properties like specific gravity and liquid limit.
• Comparing the performance of different machine learning models (ANN, SVM, GEP) for a geotechnical regression task.
• Supporting rapid decision-making in material selection for infrastructure projects in tropical regions.
• Planning soil stabilization strategies using predicted strength parameters.

Strengths

• Dataset is based on 300 physical soil samples collected from 30 distinct deposits, providing a substantive sample size.
• Laboratory testing followed ASTM standards, suggesting a controlled and reproducible data collection method.
• Three distinct machine learning models (ANN, SVM, GEP) were developed and evaluated using multiple statistical performance indices (R², RMSE, MAE, MAPE, RSE).

Limitations

• Row count is unknown, which may limit suitability assessment for specific modeling needs.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset's 68.9 KB size indicates a very limited scope, likely containing only model inputs, outputs, or summary results rather than raw experimental data.

Provenance

Source

Lateef Adamolekun via figshare.

Collection Method

300 disturbed laterite samples collected from 30 deposits and subjected to ASTM-standard laboratory testing.

Time Range

Collection date unspecified; dataset last updated May 2026.

Freshness

Last updated 2026-05 02 09:34:05; freshness should be verified.

Geography

Southwestern Nigeria.