Date of Award
Doctor of Philosophy in Business Data Science - (Ph.D.)
Michael A. Ehrlich
Stephen M. Taylor
Entrepreneurship Education researchers often measure entrepreneurial motivation of college students. It is important for stakeholders, such as policymakers and educators, to assert if entrepreneurship education can encourage students to become entrepreneurs, as well as to understand factors that influence entrepreneurial motivation. For that purpose, researchers have used different methods and instruments to measure students' entrepreneurial motivation. Most of these methods are quantitative, e.g., closed-ended surveys, whereas qualitative methods, e.g., open-ended surveys, are rarely used.
Mind maps are an attractive qualitative survey tool because they capture the individual's reflections, thoughts, and experiences. For Entrepreneurship Education, mind maps can be utilized to measure students' entrepreneurial motivation. However, qualitative analysis of mind maps in business studies has been manually performed through human coding, which is time-consuming and labor-intensive, and of questionable reliability when more than one person does the analysis.
This dissertation provides a novel automation framework to address these challenges with an interdisciplinary solution that integrates deductive and inductive qualitative content analysis approaches with the computational power of machine learning algorithms and statistical Natural Language Processing to automate the analysis of mind maps. The framework includes four sequential steps: selecting a qualitative content analysis approach, collecting and preprocessing mind maps, automating the analysis, and validating the reliability and model evaluation.
Experimentation and hypotheses testing for the automation framework are performed. The results show that the performance of classification models when applied to the automated deductive analysis of mind maps, and the performance of Structural Topic Model when applied to the automated inductive analysis of mind maps, are similar to that of manual mind maps analysis.
The utility of mind map topology in the process automation is evaluated. Findings indicate that even though inserting mind map topology as features into the dataset positively affects performance, the improvement is not statistically significant. On the other hand, treating nodes as the unit of analysis while applying Structural Topic Model to automate inductive analysis generates latent topics that follow a similar pattern to manual analysis.
This study examines the feasibility of applying the automation framework to Entrepreneurship Education research. Text classification algorithms and STM are used for the first time to automate the analysis of mind maps, and STM is applied for the first time in Entrepreneurship Education research.
The automation framework offers a unique and advanced qualitative research design that can be employed by EE researchers to benefit the EE best practices. The automation framework can enhance EE qualitative research by extracting textual statistical inference, shortening labor and time required by the analysis, and measuring entrepreneurial motivations with machine learning and Natural Language Processing techniques.
Farha, Yasser, "Mind maps and machine learning: An automation framework for qualitative research in entrepreneurship education" (2020). Dissertations. 1700.