Abstract
Machine learning techniques such as the Support Vector Machine (SVM), Random Forest (RF), M5 tree, Multiple linear regression and Artificial Neural Network was adopted to correlate soil physical parameters and California Bearing Ratio (CBR) of soils for Soaked (SCBR) and Unsoaked (USCBR). Four hundred and eighty (480) soil samples were obtained and divided into data set using training and validation of the developed models from some basic soil parameters.  Principal Component Analysis (PCA) was implemented to reduce the large dimension of the data set an the actual and predicted values from the models using Root Mean Square Error (RMSE) and coefficient of determination R2, it showed RMSE as 21.6, 21.23, 295.67, 7.03, 14.54 and 24.43,24.59,326.49,8.63,17.71 are from; MLR, ANN, MS Tree, RF, and SVM model for SCBR and USCBR values respectively. The least values 7.03 and 8.63 were observed from random forest (RF) for SCBR and USCBR. Similarly, the R values ranges between 0.1 – 0. 94 and 0.01 ─0.92 which established the relationship among the predicted and the actual SCBR and USCBR. The correlation coefficient values showed the Random Forest Model for SCBR and USCBR as the best, while the model having the least coefficient of determination R2 is the MS tree model for both SCBR and USCBR respectively.

Keywords: Modelling, Machine Learning ,  Soaked California Bearing Ratio, Unsoaked California Bearing Ratio, , Highway