Identifying Students' Pre-Classroom Behaviors in a Flipped Learning Environment

Abstract

Advancements in Information Technology have lead the world to new ways of life including in the education field. Nowadays we have various types of computer and Internet-assisted learning. With the booming of blended learning, here comes the flipped classroom environment, where students are expected to learn even before the conventional class meetings started. In this study, we address the question of how students behave toward various learning materials packaged in 3 types of media: text and images, slide shows with audio narration, and slide shows with the appearance of the lecturer. Based on our samples the findings are surprising: some students never made access before the class; and on the other hand, the text-and-image-based learning materials have the highest number of pre-classroom access.

Mixed-type Variables Clustering for Learners’ Behavior in Flipped Classroom Implementation

“…Compared to the previous works in [13]- [15], where the learners were only treated as a single population without giving any concern to the behavioral aspects that influenced either learning styles and or achievements, the clustering with mixed-type variables was found to give more meaningful insights where each cohort might be treated differently to achieve a better overall learning performance. Therefore, the results could be beneficial either for the instructors where ones might set up a learning plan where the methods and media accommodate the specific needs of the learners or for the higher education management parties where they can put students with similar interests and or behavior at the same class where the particular instructor can apply a specific treatment.…”

“…In this study, we used the same data that were used in Refs. [13]- [15]. The data came from the processed LMS log, results of the in-class quiz, and course design, as shown in Table 2.…”

“…The sequence of building the clustering model, evaluating with LGBM and SHAP, dropping features with the least impact, or modifying features were repeated several times until we get the KP5 model that is considered adequate (Figure 4). The models KM2, KP3, and KP4 were built by dropping the first access made by learners more than 72 hours after the related lecture since they are considered outliers or abnormal access [13]- [15]. However, the evaluation with LGBM to KP3 and KP4 did not give a better cross-validation score compared to KP2, therefore, in KP5 the same feature sets with KP4 were used, but the time delta included all data.…”

“…), and how these factors contribute to the learning achievements. These studies made use of questionnaires to gather feedback from students, and later, the studies in [13]- [15] augmented the data with records from the Learning Management System (LMS). LMS log has been known to provide beneficial information regarding learning activities.…”

“…Unfortunately, the students' activities that are recorded in an LMS logging system might produce a lot of data. For example, the log file used in [13]- [15] consists of more than 63 thousand records. This leads to an issue with the analysis of data with such volume [23].…”