Clickstream for learning analytics to assess students’ behavior with Scratch

Abstract

The construction of knowledge through computational practice requires to teachers a substantial amount of time and effort to evaluate programming skills, to understand and to glimpse the evolution of the students and finally to state a quantitative judgment in learning assessment. The field of learning analytics has been a common practice in research since last years due to their great possibilities in terms of learning improvement. Both, Big and Small data techniques support the analysis cycle of learning analytics and risk of students' failure prediction. Such possibilities can be a strong positive contribution to the field of computational practice such as programming. Our main objective was to help teachers in their assessments through to make those possibilities effective. Thus, we have developed a functional solution to categorize and understand students' behavior in programming activities based in Scratch. Through collection and analysis of data generated by students' clicks in Scratch, we proceed to execute both exploratory and predictive analytics to detect patterns in students' behavior when developing solutions for assignments. We concluded that resultant taxonomy could help teachers to better support their students by giving real-time quality feedback and act before students deliver incorrectly or at least incomplete tasks.

Identification of the Students Learning Process During Education Robotics Activities

“…This is a similar result compared to Blikstein et al (2014), who identified that a "steadier incremental steps" strategy of programming correlated to a better performance in the resolution of the exercise. Pathway 4, with the highest number of trials (57) ( Table 2), contains teams that did not obtain a working sequence in their first part of their work, and this result is similar to Chao (2016) but opposed to Filvà et al (2019).…”

“…However, new data mining and machine learning technologies allow researchers to capture detailed data related to problem-solving and programming trajectory of a large number of learners . Recent studies (Berland et al, 2013;Blikstein et al, 2014;Chao, 2016;Wang et al, 2017;Bey et al, 2019;Filvà et al, 2019) have mostly applied machine learning techniques to data gathered from students during programming activities without the presence of physical robots, obtaining good results in the identification of different patterns in specific coding tasks ( Table 1 summarizes machine learning techniques and features selected in these studies). Berland et al (2013) and Chao (2016) used a k-means algorithm to discover patterns in the programming activity of novice programmers; the first study identified three general patterns (Tinkering, Exploring, and Refining) and presented a positive correlation between the quality of the programming sequences designed by the students and two of the emerged patterns (Tinkering and Refine).…”

Systematic Review on Which Analytics and Learning Methodologies Are Applied in Primary and Secondary Education in the Learning of Robotics Sensors

“…Projects foster an environment of discussion, creativity, problem-solving, inquiry, modeling, and testing, and may be applied to all grade levels and subjects, such as programming. For instance, combining ER with visual programming makes it more attractive, fosters students’ attention and interest, and results in immediate feedback [ 64 , 65 , 66 , 67 ].…”

Personal Data Broker Instead of Blockchain for Students’ Data Privacy Assurance

“…Learning Analytics has become a key tool for assessment [1]. The students' interactions within online learning environments are collected, processed by statistical models, and finally presented to teachers and other stakeholders.…”