Leveraging Artificial Intelligence And Statistical Methods For Evaluating Geotechnical Properties Of Lateritic Soil In Road Construction: A Case Study On The Umuahia-Ikot Ekpene Road

Abstract

Geotechnical properties of lateritic soils are critical for road construction, influencing subgrade stability and pavement performance. This study investigates these properties along the Umuahia–Ikot Ekpene Road, with a focus on predictive modeling and variable importance. Four locations were sampled at depths of 150 mm, 300 mm, and 450 mm. Key properties, including moisture content, particle size distribution, Atterberg limits, compaction characteristics, and California Bearing Ratio (CBR), were analyzed. Moisture content ranged from 10.8% to 13.0%, with Locations A (12.2%) and C (12.8%) exhibiting the highest values. All soils were classified as well-graded, with gravel content varying from 12.5% to 16.0%. Atterberg limits indicated medium plasticity, with plasticity index values between 12% and 14%. Maximum dry density (MDD) ranged from 1795 to 1850 kg/m³, Optimum Moisture Content (OMC) from 10.8% to 12.1%, and degree of compaction from 95% to 99%. Unsoaked CBR values were 24-31%, and soaked CBR values were 17-23%. Lasso Regression outperformed other models (Linear regression, ridge regression, decision tree regressor, gradient boosting regressor, support vector regressor and random forest regressor) in predicting CBR, achieving R² values of 0.86 (unsoaked) and 0.98 (soaked). Feature importance analysis highlighted the key soil properties driving CBR variations while sensitivity analysis reveals that MDD (kg/m³) has a significantly greater impact on the predicted soaked CBR values compared to moisture content (%). The results suggest that these lateritic soils possess moderate strength, making them potentially suitable for subgrade materials. However, stabilization is likely required for sub-base or base layers in road construction.