In geotechnical engineering, determination of fine grain soil properties plays a crucial role for all the projects and requires considerable amount of time and effort. Due to technological advancements, nowadays suitability of soils for a project is ascertained using different traditional and AI-based models in the preliminary stage. Generally, prediction of fine-grained soil properties using a traditional regression model have difficulty due to their complexity and nonlinearity. Hence, the present study has carried out using artificial neural networks in assessing the soaked CBR and compacting behavior of fine-grained soils. The laboratory-measured data was used to compare the performance of ANNs with that of traditional regression models. The results have shown a clear superiority of ANNs against regression models, especially at small data sizes. The models of neural networks are able not only to provide more accurate predictions of the properties of soil but to discover subtle interdependencies between the parameters of soil that can hardly be caught by the traditional methods.