The graphic well log offers a quick sight view of the geological condition and underground facies. If it isn't viable to access the graphic well log, machine learning techniques may be used to predict facies from well logs or cores. In this regard, the class imbalance venture makes it difficult for classifiers to adequately learn and distinguish among minority and majority classes. This research focuses on the development of a cost-sensitive classifier by modifying the objective function of the well-known random forest algorithm to construct a synthetic graphic well log. First, a standard version of the algorithm was implemented to provide a basis for performance comparison. In the second stage, a cost-sensitive algorithm was developed. The results indicate that the cost-sensitive method performs better than the standard algorithm in this case study.