Prediction of Soaked CBR Value of Sub-base Soil Using Artificial Intelligence Model

The objective of the research is to develop a predictive model for evaluating the California bearing ratio (CBR) value of soaked soil by using conventional and hybrid artificial intelligence models. The study used field soil samples from a highway construction project area and gathered relevant input values based on literature recommendations and data analysis. The research aims to create reliable and simple predictive models employing artificial neural networks (ANNs) with regression analysis (RA) based on soil features such as gradation, Atterberg limits, and compaction qualities. A database comprising 197 CBR values from quality control reports of the Mid-Hill Road construction project in Nepal was compiled. The building of the model used around 70% of the data, while the validation of the model used about 30% of the data. Both RA and ANN were employed and evaluated for their prediction accuracy using the coefficient of determination (R2). The research mainly focuses on the importance of computational modeling in CBR value prediction and presents a comprehensive comparison between conventional and hybrid AI models. The findings bear significant implications for advancements in soil testing for sub-base soil applications.