Borehole logs are essential for gaining insights into the subsurface conditions of a site, as they offer a comprehensive characterization of the different soil layers at varying depths of the site. Estimation of crucial subsoil parameters through borehole profiling at a site enables us to explore various elements of civil foundation planning. However, spatial and temporal variability in soil characteristics of a site leads to high-density borehole sampling requirements for the precise estimation of soil parameters, which is both resource-intensive and time-consuming. There is ample scope of leveraging advanced machine learning techniques to make depthwise estimation of subsoil parameters at different locations (termed as missing borehole locations) using borehole data from other locations within a given site. In view of this promising scope, a machine learning-based framework is proposed in this study to explore the possibility of effective subsoil profiling with low-density borehole sampling. In particular, this article intends to present a study on the variation of the cohesion of soil as a crucial shear strength parameter across different depths as well as locations of boreholes and explore the applicability of various machine learning-based regression models to predict the cohesion for different depths at a missing borehole location as accurately as possible. The experimental results indicate that the soil cohesion parameter of the borehole profile assumed to be missing at a location during experimentation can be predicted by the ANN-based convolutional regression model with a promising accuracy (86% R2 score) while validating with the original values obtained from the laboratory test results based on the actual borehole logs at the presumably missing location.