Abstract:

Technological innovations have modified the traditional learning approaches and thus have strengthened the educational system. In the conventional classrooms, teachers monitor students’ learning processes; they analyze their performance through observation. Computer-Based Educational Systems is developed under artificial intelligence approaches being used in classrooms in many countries around the world. In this paper, the information obtained from the interaction between students from Federal No. 2 of the city of Xalapa, Veracruz, México, with the intelligent tutor “Scooter” is analyzed with the objective of modelling the learning acquired by students. Each student’s interactions are stored based on a time sequence. In addition, the number registered for each student is not the same because each student interacts more or less times with the ITS. It is of great importance to analyze the large quantity of information that students generate when interacting with Scooter when solving an exercise. The results suggest that educational data mining allowed the discovery of variables that mostly represent the students, which are: success when doing an activity, learning probability, type of answer, and finally, registered time. Applying clustering algorithms it was determined that students must be grouped in 3 groups of data. The clusters are associated to the learning level that students acquire when solving an exercise in the mathematics intelligent tutor. The students profiles identified and associated to learning are categorized as: lacking learning, satisfactory learning, and advanced learning. Once the groups of students’ profiles according to their learning are obtained, a predictive model is defined for the decision making. The predictive model definition is based on the Random Forest algorithm (decision tree) which allows the identification of students that show more advancement, as well as students with difficulties to solve the exercises that were presented. Based on the Random Forest tree the rules IF-THEN are inferred, which allow defining a student profile before interacting with the Scooter Intelligent Tutor System. Finally, Educational Data Mining allows reaching the conclusion that the characteristics found are useful to the professors to help in the teaching- learning process; due to the fact that, once students are grouped in the classes: lacking, satisfactory, and advances; the model will allow students, other support staff, parents, and even students, to get to know how they learn while using the tutor. This will allow the identification of students that need reinforcement and feedback in the most complex mathematics topics which have an effect in decreasing the failing rate in the subjects related to mathematics.

Keywords:

Intelligent tutoring system, educational data mining, learning, pattern recognition.