Data Visualization and Reporting
Randyll Pandohie
PhD Student
University of Central Florida, United States
Haiyan Bai, Ph.D.
Professor
University of Central Florida, United States
Haiyan Bai, Ph.D.
Professor
University of Central Florida, United States
Location: Grand Ballroom 10
Abstract Information: Categorial variables in evaluation research has progressively increased in recent years with survey and other types of data (e.g., media postings). This presentation will introduce and discuss about the concepts dealing with categorical variables and explores machine learning techniques for statistical inferences using these types of variables with a focus on empowering story telling in evaluation. The use of categorical variables is demonstrated in the education space by investigating the features that contribute to learning such as gender, family size, parents’ higher education. In this presentation, a reinforcement machine learning model will be presented using the categorial variable transformation procedures. This provides a novel approach to the evaluation and measurement of variables that traditionally excluded from statistical models through utilizing machine learning to create a reinforcement model. By using the features of machine learning we are able to give an accurate representation and improve the communication and impact of interventions.
Relevance Statement: This proposal presentation will help evaluators to learn how to use machine learning as a powerful tool to help with their storytelling for understanding and communicating the impact of programs and interventions. By analyzing large amounts of data, machine learning can help evaluators identify patterns and insights that may be difficult to detect through traditional evaluation methods. One application of machine learning in program evaluation is in the analysis of unstructured data, such as social media posts, online reviews, and open-ended survey responses. Machine learning algorithms can be used to identify themes and sentiments in this data, providing insights into how program participants perceive the program and what aspects of the program are most effective or in need of improvement. Developing an understanding of the variables that contribute to student performance (Cortez 2008) - such as demographics, gender, family size, study time, absence, social factors, online learning access – will play a vital role in determining their impact on the resources and strategies used for learning (Afzaal, 2021). The use of many categorical variables in these data leads to difficulties in interpretation of results. How to handle categorical variables in models with high dimensions is an open research area (Kuhn 2019). Reinforcement Machine Learning, which models the interaction with the environment, can be utilized to learn based on a series of actions. More specifically, the following research questions are to be addressed in this presentation: 1. How can we develop statistical models for the harnessing of categorical information in evaluation? 2. What are the typical variables that influence student performance in the information age? 3. How can reinforcement machine learning be utilized for data exploration in the education data domain? Another application is in the analysis of program performance data. Machine learning algorithms can be used to identify correlations and causal relationships between program inputs, outputs, and outcomes, providing insights into what factors contribute to program success or failure. This analysis can help evaluators identify opportunities to optimize program design and implementation, leading to more effective interventions. Overall, the power of storytelling using machine learning for evaluators lies in its ability to provide insights into program impact that may be difficult to detect through traditional evaluation methods. By analyzing large amounts of data and identifying patterns and insights, machine learning can help evaluators communicate the impact of programs and interventions more effectively, leading to more informed decision-making and better outcomes for program participants. The expected outcome of this research will provide the basis for treating categorical variables in models for evaluation. This will provide a blueprint for the use of categorical variables in other types of modeling, which will lead to an improvement in the prediction and understanding of these features. The use of this approach in the education domain provides an application foundation to test this categorical framework in various other data spaces in evaluation. The creation of an application in machine learning provides application in computer science for the reinforcement methods to improve validity of evaluation with variables.
Presenter: Haiyan Bai, Ph.D. – University of Central Florida
Presenter: Randyll Pandohie – University of Central Florida