Abstract In the educational environment field of teaching and learning for the SQL language, there have been recent approaches that use heuristic methods to categorize and discover information about student errors. The categorization of student source code errors by the level of complexity of the problem was one of the limitations found in these approaches. Owing to this limitation, the conclusions drawn from these studies are superficial and do not explain the real reasons that led the students to make their mistakes. As an alternative, we propose a new methodology that uses cluster-based decision trees and the fundamentals of the programming language to discover the most relevant patterns of knowledge about student behaviour. Through an innovative architecture for classifying errors, the methodology allows us to explain: (a) what are the most frequently committed errors, (b) when these errors usually occur and (c) the reasons that led to the error. The methodology was applied in a case study with more than 105,200 SQL queries, which, in addition to explaining the hypotheses about what leads students to make mistakes, also gave the following results: (1) it proved the concept of reducing the breadth of analysis of the decision tree, (2) it showed that an error database could be used for the discovery of knowledge and (3) it resulted in the discovery of a set of 66 rules; 12 related to semantic errors and 54 to syntax errors. The rules identified the error, when the mistake was made and why the student made the mistake. The methodology uses an architecture based on the fundamentals of the programming language and can therefore be adapted for use in another context.
Paper
Full text
Application of clustering-based decision tree approach in SQL query error database
Semantic Scholar · Computer Science · 2019
Abstract
Abstract In the educational environment field of teaching and learning for the SQL language, there have been recent approaches that use heuristic methods to categorize and discover information about student errors. The categorization of student source code errors by the level of complexity of the problem was one of the limitations found in these approaches. Owing to this limitation, the conclusions drawn from these studies are superficial and do not explain the real reasons that led the students to make their mistakes. As an alternative, we propose a new methodology that uses cluster-based decision trees and the fundamentals of the programming language to discover the most relevant patterns of knowledge about student behaviour. Through an innovative architecture for classifying errors, the methodology allows us to explain: (a) what are the most frequently committed errors, (b) when these errors usually occur and (c) the reasons that led to the error. The methodology was applied in a case study with more than 105,200 SQL queries, which, in addition to explaining the hypotheses about what leads students to make mistakes, also gave the following results: (1) it proved the concept of reducing the breadth of analysis of the decision tree, (2) it showed that an error database could be used for the discovery of knowledge and (3) it resulted in the discovery of a set of 66 rules; 12 related to semantic errors and 54 to syntax errors. The rules identified the error, when the mistake was made and why the student made the mistake. The methodology uses an architecture based on the fundamentals of the programming language and can therefore be adapted for use in another context.