Unfolding the Drivers of Student Success in Answering Multiple-Choice Questions About Microsoft Excel

Abstract

Many university programs include Microsoft Excel courses given their value as a scientific and technical tool. However, evaluating what is effectively learned by students is a challenging task. Considering multiple-choice written exams are a standard evaluation format, this study aimed to uncover the features influencing students’ success in answering these types of questions. The empirical experiments were based on Excel evaluation exams containing questions answered by 526 students between 2012 and 2016, with a total of 3,340 answers characterized by 17 features. Through data mining, a neural network was developed that accurately modeled students’ choices. A sensitivity analysis was applied to the model to assess the most relevant features. Findings identified four highly relevant features for students’ success: number of words of the question, topic, difficulty degree, and number of similar choices. This study helps to guide the design of future exams by quantifying the individual influence of each feature.

Keywords:

• Data mining
• Excel
• feature relevance
• multiple-choice questions
• students’ performance