Poster 7

June Yu

Texas State Computer Science

Identifying Resilience Factors in Texas Public Schools

Abstract: Decisions about when and how to reopen schools were difficult for district administrations during COVID-19, as there was no consensus on the impact of reopening of the school reopening on the spread of COVID-19. Learning loss was documented throughout the process. In this project, we attempt to identify the most impactful factors on learning loss (or the absence of learning gain) at the Texas level and help policy makers make more informative decisions on learning recovery by creating a dashboard of our analytics and the most impactful factors. We analyzed all school districts in Texas and pull data from the average STAAR test results for math and reading subjects to create a label, Learning Loss, Expected, and Learning Gain. The data have been integrated from different sources such as grade, race/ethnicity, and Title 1/free lunches from NCES Common Core of Data, local unemployment from U.S. Bureau of Labor Statistics, in-school student population on 9/20, 10/20, and 01/21 from Texas Health and Human Services, demographic data from Census, and emergency relief grant programs for schools from Texas Education Agency. Our initial EDA showed that the negative impact of COVID-19 erased years of improvement in reading and math between 2019 and 2021. 90 predictors and their importance were evaluated using 9 different feature selection methods to identify the most impactful predictors to predict the learning loss. We also built the state-of-the-art Gradient Boosting models to examine the influence of those impactful predictors.