Borehole geophysical log data of basins are essential in mining and hydrocarbon exploration as a relatively low-cost and accurate method for subsurface lithological examination. They provide estimations about the extent and quality of minerals, including the coal content of the basin. The well-log data usually contain continuous density, acoustic, conductivity, gamma ray, and neutron density measurements from which the pattern of lithology can be explored. A conventional M-N cross-plot technique has been used for well-log data analyses. However, over the past decade machine learning algorithms have been increasingly employed in different fields of sciences and engineering, including in geosciences. In this study, we train machine learning estimator models for coal evaluation required for mining operations using well-log and core data. This new method is expected to provide better, fast, and more reliable predictions than the traditional analysis. This study uses the geophysical well-log measurements of open-source data of the Athabasca Oilsands basin. Our initial estimations are based on 100 well-log data (Gamma-ray, neutron, density, resistivity, and caliper as features) and their corresponding core data as class labels. Deep and Random Forest estimator models reveal 76% and 86% prediction accuracies, respectively, on unseen test data.