Leveraging Software for Predictive Analytics George Gonzlez Director

Leveraging Software for Predictive Analytics George González, Director of Institutional Research & Effectiveness Michelle Callaway, Manager of Program Review San Jacinto College, Pasadena, Texas

San Jacinto College § Three campuses in Pasadena and Houston, Texas § Fall 2017: 30, 509 credit students § AY 2016 -17: 7, 500 credentials awarded (Associates & Certificates) § 2017: Aspen Institute Rising Star Award – Top 5 Community College in the Nation

History § 2011: Initial logistic regression models built using Base SAS programming to predict FTIC fall to spring persistence § Limitations: § Extensive Coding = Time Consuming § Doesn’t lend itself to collaboration § Editing the model requires re-parameterization § At. D Data Coach, Dr. Jing Luan, suggested SAS Enterprise Miner

Prep Work for Model Building § Strong partnership with ITS § Data included in Model § Demographic data § High School Course Taking Behavior § Financial Aid data § Dual Credit data § College readiness level § Outcome variables

SAS Enterprise Miner § 2014: Used SAS Enterprise Miner to build decision-tree models for predicting FTIC fall to spring persistence § Benefits: § § § Less time coding = More time for analysis Easy to collaborate Build models in-house Editing models is relatively quick Low cost: SAS Enterprise Miner ~ $1500/license/year; Base SAS programming software ~ $800/license/year § Used results to identify our most at-risk students

Currently Building § Model to identify FTIC students at risk of failing or withdrawing from all courses in first term § Math pathways models to identify best math course for FTIC student success § Interactive tool that incorporates model results with student information system (SIS) data

SAS Enterprise Miner Workflow Example

Future Work § Using predictive modeling to identify optimal combinations of student interventions § Incorporating non-cognitive measures into models

Questions?