Applying artificial intelligence data mining tools to the challenges of program evaluation

Educational data warehouses, particularly in this accountability age of No Child Left Behind, are replete with a variety data. The data represents a wide range of variables (input, output and process). Up until now, educational leaders have not been able to analyze the variety and plethora of data they have at hand to answer complex questions of root cause. The data warehouse in a suburban, New England K–8 school district was analyzed using Artificial Intelligence data mining tools. These tools have proved useful and beneficial in the business world for many years. The superintendent, like most educational leaders, asked a predictive question regarding the efficacy of a newly implemented reading program in the lower elementary grades. First, a traditional analysis was conducted on first and second grade data which determined that students who had received a new reading program performed statistically significantly better than those who did not receive the new program. Next, through the use of a classification and regression tree (CART) algorithm, the lower elementary data was successfully mined. From an extended selection of second grade variables (reading program, writing sample scores, gender, attendance, parent/guardian, ethnicity, kindergarten experience, readiness screening and teacher), the CART algorithm identified the reading program variable as the most important variable in relation to the target variable (Gates Reading Test comprehension section). This result, generalized to the district studied, indicated that the reading program variable was predictive of achievement on the target variable (Gates Reading Test comprehension section). This analysis provided meaningful and thought-provoking information to the superintendent and assistant superintendent regarding their reading programs. Data mining analyses can now be conducted over the course of a few days. The analyses also stirred questions of which program variables are predictive of other target variables preK–8. The application of Artificial Intelligence data mining tools on educational data warehouses could enhance, supplement or even fast-track traditional program evaluation processes for educational leaders.