Text Classification to Detect Student Level of Understanding in Prior Knowledge Activation Process

The utilization of Intelligence Tutoring System is expected to improve the capability of Self-Regulated Learning (SRL), aiming to allow the student to understand what becomes their objective of the study and allow to apply the appropriate strategy to achieve the purpose of the study. Prior Knowledge Activation (PKA) is the part of Intelligence Tutoring System which aims to detect the level of understanding from student through the pretest. The pretest aims to recalling the knowledge about the topic which will be learned. The result from student paragraph citation will be grouped into high, medium, and low. Grouping is used to reflect the knowledge from student through the learning objective and determine the sub goal of material which will be studied. This research proposes the weighting word method combined with seven machine learning algorithms to compare the best algorithm in paragraph text citation, with 40 data sets from student paragraph text. The results show that Logistic Regression and Multi-Layer Perceptron Algorithms give the same accuracy and kappa which are 90.19% and 0,84 respectively. However, both consume time differently in performing classification. Thus, algorithm which is more featured is Logistic Regression Algorithm which consume 0.33 s while the Multilayer Perceptron consume 10, 12 s for the classification process.