Education

Education


Increasing High School Graduation Rates: Early Warnings and Predictive Systems

Roughly one in five students in the U.S. – nearly 700,000 students each year – do not complete high school on time. To help more students graduate on time, school districts across the country use intervention programs to help struggling students get back on track academically. Yet in order to best apply those programs, schools need to identify off-track students as early as possible and enroll them in the most appropriate intervention. Increasingly, forward-looking school districts are exploring data-driven “early warning systems” that can help schools find students in need of extra support. These current identification systems are hyper-specific to each district and rely mostly on anecdotal evidence or teacher intuition rather than data-driven tools, which means that resources are not allocated most effectively.

By working with diverse school districts across the country, DSaPP is using modern data science and machine learning methods to create a robust set of predictive models that improve upon current identification systems in three main ways:

  • Allows school districts to prioritize interventions. Our approach outputs a ranking of students, giving educators the ability to maximize their limited resources by targeting the students most at risk of dropping out first.
  • Scales across multiple school districts. Our open source models are deployable across multiple school systems with limited maintenance needs, meaning school districts do not need to start from scratch to build an in-house system.
  • Approaches the problem flexibly and adaptively: Our models can accept and integrate a wide range of data easily. Additionally, they can update as cohorts graduate to capture changing trends.

School districts and nonprofit partners can use the informational outputs from these models to build internal dashboards that accurately assess individual students’ needs in order to increase their chances of graduating on time.

We are currently working with the Wake County Public School System and the Cabarrus County and Kannapolis City school districts in North Carolina, the Arlington Public School District in Virginia, and Vancouver Public Schools in Washington in order to develop and refine our system.

To read more on the initial stages of this project, please see the 2015 DSSG blogpost on Identifying High School Students Who May Not Graduate on Time.

Predicting College Persistence and Targeting Support

Low-income students disproportionately leave college before completing a degree. DSaPP is building a model to predict which students are most at risk of not persisting through college to help schools ensure that their students and alumni receive the support they need in order to graduate.

In order to build this model, DSaPP is working with KIPP: Chicago, KIPP: New Jersey, The NOBEL Network, and Perspectives Charter Schools. High-performing charter schools graduate students from low-income backgrounds at high rates. Many of these students begin college, but most never complete their degrees. In order to improve their alumni’s college completion, many charter networks have alumni counselors who provide alumni support throughout college, including academic and financial advising. These alumni counselors have limited time and resources to reach out to every student in need and would benefit from prioritizing its service to students at greater risks earlier rather than later.

By working with several charter school networks across the country, DSaPP is using modern data science and machine learning methods to create a robust set of predictive models that can aid these alumni counselors in three main ways:

  • Allows charter schools to prioritize interventions. Our approach outputs a ranking of students, giving alumni counselors the ability to maximize their limited resources by targeting the students most at risk of dropping out of college first.
  • Scales across multiple charter schools. Our open source models are deployable across multiple school systems with limited maintenance needs, meaning charter schools do not need to start from scratch to build an in-house system.
  • Approaches the problem flexibly and adaptively: Our models can accept and integrate a wide range of data easily. Additionally, they can update as cohorts graduate to capture changing trends.

Charter school partners can use the informational outputs from these models to build internal dashboards that accurately assess individual students’ needs in order to increase their chances of graduating college.

We’re looking to hire faculty members (at all levels) interested in working at the intersection of data science and public policy. Contact us if you’re interested in learning more about the positions.

Our initial findings are available here.