Soil erosion by water and other anthropogenic activities in the semi-arid and humid region is noticed as a major issue in the reduction in natural land by the loss of soil nutrients. The seven standard parameters were suggested in the literature for the assessment of soil erosion hazard, viz. soil loss, sediment yield, run-off potential, land capability class, drainage density, sediment transport index, and slope. In the present study, the combination of intelligent vulnerability prediction, multi-criteria decision-making, and geographic information system techniques provides an effective approach to identify the soil erosion hazard in the semi-arid and humid region. It makes this process more effective and efficient as the vulnerability of soil erosion hazard can be predicted by the proposed trained models for any locations that have the streamlined values of above seven parameters as suggested in this paper. The standard machine learning classifiers such as k-nearest neighbour, decision tree, random forest (RF), multinomial naive bays, adaptive boosting, and gradient adaptive boosting (GAB) have been applied on the spatial data set of “Pairi” river watershed found in “Chhattisgarh”, India. There are five categories of soil abrasion, viz. “very low”, “low”, “medium”, “high”, and “very high”, in this data set that represents an index of soil erosion hazard. The experimental results have given 91.5140% and 90.5525% accuracy using RF and GAB, respectively, whereas a much better log-loss measure, i.e. 0.27, is obtained by the GAB in comparison of 0.93 with RF. The results have been verified by visiting the ground truth locations.