Tinkering Behavior Detector Using Multiple Linear Regression: Development of Intelligent Tutoring System for Novice C Programmers

Coding regardless programming languages is an expected skill evaluated in computing courses [1]. The majority of programming research has been on student challenges and mistakes that compromise technical accuracy [2]. Prior study has shown that debugging approaches are utilized to create response in learning coding [3]. The goal of this paper is to build a web-based intelligent tutoring system (ITS) application for freshmen learning C language that predicts tinkering activity using constraint-based modeling (CBM) and machine learning algorithm. The system was created for capturing C code samples focused on assignment statements, which would then be sent to the tutor model for submitted code assessment and response generation using CBM. This paper expanded the first system prototype of Arguson et al. [4] which was pilot tested on 2 synchronous coding classes among 31 freshman tertiary students composed of 9 BS Computer Science and 22 BS Information Technology respondents. The dataset was taken from the aforementioned first prototype, whilst the student model was built using the data science approach for building the student model. Arguson et al. [5]’s study was used to construct a modified CBM that centers on assignment statements. The model statistically predicted tinkering on both numerical value problems and incorrect expressions performed by beginner programmers, according to the results. In the development of the final ITS prototype, a novel detection of tinkering activity in the context of C coding focused on assignment statements was deployed. Adhering to Scrum framework, the authors were able to supervise the software project in this study. It emerged that tinkering activity can be predicted using compiled inexperienced C programming logs as well as tailored-fit feedback is required to comprehend the aforementioned programming language.