Abstract: Mass open online courses (MОOС) are an example of the development of the open learning movement, which has attracted a lot of attention from both the academic and public spheres. IOC is not an independent phenomenon isolated from other developments in the field of open and distance learning or educational technologies. On the contrary, IOCs are closely linked to other developments in this field, have the potential to support lifelong learning, remove barriers to learning, ensure equal opportunities in education and, most importantly, liberalize knowledge.

The theoretical bases of formation of the market of Internet courses are defined in the work; the current state and trends of the Internet courses market in Ukraine and the world are analyzed; the classification of Internet courses depending on their rating assessment is carried out; the rating value is predicted to determine the popularity of online courses. The solution to the problem of predicting the popularity of courses in this study is achieved through machine learning methods that classify online courses based on the rating parameter. Namely, the courses that receive the maximum rating score are considered popular. Solving problems of classification or regression by machine learning is most often achieved by building 171 ensemble models. This approach is based on the hypothesis of combining several models, which could lead to the formation of a more powerful model. The method of combining models should be adapted to their types. Currently, there are several meta-algorithms used to form integrated models. In one of them (the method of bugging) homogeneous initial models are studied in parallel and independently of each other, and then combined according to a certain deterministic averaging rule. Currently, there are several meta-algorithms used to form integrated models. In one of them (the method of bugging) homogeneous initial models are studied in parallel and independently of each other, and then combined according to a certain determined averaging rule. One variant of this algorithm is the random forest method. In another algorithm, models are trained sequentially in an adaptive manner. The most popular of these are adaptive and gradient boosting. The first updates the weight of each of the training dataset objects, and the second updates the values of these objects. In doing so, both methods attempt to solve the optimization problem to find the best model represented by the weighted sum of the initial weaker models. In this paper, gradient boosting and random forest algorithms are used to predict the popularity of online courses. The proposed models guarantee 65 percent accuracy of forecasts. Factors that reduce the accuracy of the forecast include attributes that do not correlate much with the forecast value, as well as the disparity and significant emissions observed in the data. The considered methods of machine learning are subject to modifications and tuning, which makes it possible to improve the modeling of the classifier