As a result of high oil price fluctuations, the challenges in optimizing drilling operations are increasing. This costly operation is considered as the core of the oil and gas industry, and there is a need for continuous monitoring to reduce its costs. In well design, the key to successfully achieving objectives is to design well programs based on the prediction of potential problems. The most common of these problems include inefficiency of the bottomhole assembly during drilling, hole pack-off, pipe sticking, mud loss, fall of penetration rate, well deviation, tight spots in the well, pipe failure, mud contamination, damage to the formation, well cleaning problems, and problems related to equipment and staff. Understanding and predicting drilling problems, understanding the causes, and planning solutions are essential to control well costs. The ever-increasing complexity of wells has exacerbated drilling costs and driven the drilling industry's focus on reducing non-productive and lost times. Determining the critical depths where drilling problems are more likely than other depths can play a key role in predicting and reducing the probability of encountering these problems. The goal of this research is to use machine learning methods to identify and classify drilling problems using drilling parameters. Drilling data recorded during the drilling of 17 wells in one of the offshore fields in southwestern Iran were used to build machine learning models. 411 data sets were extracted after removing outliers and unsuitable samples from various sources such as daily drilling reports, daily drilling mud reports and well final reports and were used to train the models. Information including various drilling parameters and problems of each depth (if any) was obtained from the reports. Also, by applying feature selection methods, features was reduced from 31 to 14 features. Then, using MATLAB R2021b software and 75% of the data, multi-class classification models including dec