This doctoral thesis describes a research whose main objective was to develop a model for the evaluation of learning outcomes in STEM education based on the principles of educational recommendation systems. The developed model made it possible to connect the data collected in the system during the assessment of students' knowledge with the data of the educational recommendation system, all with the aim of promoting students' motivation to learn and generating different types of recommendations that enable students to achieve learning outcomes. Learning outcomes are written statements that define what students should know, understand and/or be able to demonstrate upon completion of a particular learning process. Knowing and actively using learning outcomes enables students to organize time and planned learning activities optimally and creates a basis for better learning results. During learning, students can be continuously supported by educational recommendation systems. Their main purpose is to generate appropriate recommendations for individual students based on algorithms developed for their work. On the other hand, online evaluation systems represent a tool that can be used for formative and summative evaluation while providing quick feedback on the success of students' work. The model developed in the conducted research combines the functionalities of these two systems, with learning outcomes serving as the connecting element. In STEM education, solving math tasks is one of the most common ways to assess students' knowledge. Most of the systems used today are based on checking the accuracy of the final outcome, while the data on the accuracy of intermediate results in these tasks is ignored. Monitoring the accuracy of intermediate results can enable continuous monitoring of the achievement of concepts, i.e. the achievement of learning outcomes, while providing timely recommendations to students to positively influence their motivation for further work. The system model developed in this research includes the monitoring of students' work success at the level of intermediate results in the tasks, which is one of its distinctive features. The developed model enabled the successful implementation of a continuous online evaluation in the educational process. The observation of student performance in continuous knowledge assessment at the level of intermediate results in solved tasks enabled the creation of an extended data set for system operation. This extended data set served as a basis for monitoring the current level at which students have mastered concepts, i.e. learning outcomes, and for creating three types of individualized recommendations: recommendations of associates, recommendations for the next learning step, and recommendations to increase motivation. The above recommendations had a positive impact on student motivation which led to better results of learning activities during the semester and enabled students to achieve the learning outcomes. The designed model was implemented in the form of a network application, the functionality of which was tested in work with students enrolled in university and vocational undergraduate study programs at the Faculty of Engineering, University of Rijeka. Algorithms based on the application of fuzzy logic were developed for the operation of the network application, using an extended data set collected by conducting a continuous online evaluation of students knowledge. The research was conducted using the design-based research methodology, and in the experimental phase of the research, six cycles were conducted in five consecutive years. The results of the conducted research confirmed the positive impact of the developed model on the motivation of the students involved in the research, the success of the model in helping students achieve learning outcomes based on the recommendations generated for each student, and the achievement of better results of learning activities during the semester.