Predictive Modeling a Tool for Student Retention

Pam Roman

Share this article

If you’ve ever stepped on a bathroom scale, you’re familiar with – at the most basic level – the concept of collecting and using data to affect behavior. Based on our experiences and personal knowledge, we can make the connection between eating too much pizza and what the scale will tell us.

But what happens when enterprises want to understand their customers’ behavior? How do we derive the connections and find the answers to affect the positive outcomes we desire when the behaviors become so complex, with so many more variables and other considerations? That’s what our solution – Predictive Behavior-Based Value Modeling – is for. It uses artificial intelligence (AI) and sophisticated data analytics to look at behaviors and data, make connections among naturally occurring “digital footprints,” and generate valuable, actionable information better and faster than a human ever could.

Our Client

Wichita State University (WSU), a public university, is the only metropolitan university in the Kansas Board of Regents’ system. The University’s 14,000 students study business, education, engineering, fine arts, health professions, and liberal arts and sciences at the undergraduate and graduate levels. The University’s vision statement includes being known for providing impactful student experiences and driving prosperity for the people and communities the schools serves.

Challenge

WSU (and the educational system in general) has wrestled with the continuing impacts of COVID-19. Once thought to be a short-term crisis to be endured until normalcy returned, the pandemic has evolved into a longer-term threat to overall student success. Time away from campus and the newness of all online classes increased the risk of students falling back or dropping out.

The University recognized this issue and, in line with their vision, sought a more proactive way to help ensure their students succeed both at school and beyond. The school wanted a way to identify “at risk” students – students who were likely to either drop out or not graduate on time. But the normal methods to identify these students (such as grades and attendance) are too reactive in nature. It takes time for grades to get into the system, so by the time an “at risk” student is identified under these methods, precious recovery time has been lost.

Instead, WSU wanted to predict which students are likely to become at risk. This would enable them to provide earlier intervention and keep those students on track before it became harder to help them.

Action

The Prolifics Predictive Behavior-Based Value Modeling solution is composed of the following specific steps. The solution leverages third party data and AI offerings, as well as Prolifics accelerators for data and AI, and other Prolifics IP assets and services.

We applied these steps with at WSU as follows:

  • Prolifics Predictive Models Workshop
    • During this workshop, we walk through specific predictive use cases to determine the scope and data readiness to achieve the desired models. There was significant work necessary to prepare the client data. The University suffered from data silos and a myriad of different running tools that hampered its ability to effectively manage its data.
    • WSU also needed to include its community college network as part of this initiative. University students often take classes across this network, so access to that information was an important piece of the puzzle. In addition, the school wanted to incorporate student activity from the community. The University placed a heavy emphasis on community involvement as part of the overall educational experience, so they needed real time access to data from their community partners.
  • Deploy Data Platform
    • Although represented as a “detour” on the above chart, at least 90 percent of our clients must execute this step and the one that follows. Often the initial hurdle is that client data is trapped in older, legacy and siloed systems. This was the case with our client. To centralize their data access, Prolifics deploys a centralized data platform to serve as the hub for all data feeding the predictive models.
  • Migrate and Cleanse Data
    • To resolve the University’s internal data silos into the new centralized data catalog, Prolifics migrated their current data and implemented a managed approach to the data governance, business terms and their data glossary.
  • Build Predictive Models
    • Prolifics built models around predicting student success and student risk. These models leveraged specific real-time indicators to predict how well students would perform; it also indicated where instructors or counselors will need to give extra attention to students who may be struggling.
  • Build Visualization and Alerting
    • We created and deployed a customized dashboard with alerting capability. This gives WSU the easy, immediate access to the information needed to make real-time decisions for intervention needed for their students, to ensure they stay on track.

Results

This solution is scheduled to go live January 2021, so production metrics are not yet available. The University will use the following metrics for improving student success:

• Reduced dropout rate (how many students drop out before graduating)
• Increased on time graduation rate (how many students graduate within 4 years)
• Improved student satisfaction scores

In addition to caring about students and their success, the University’s achievement of the above metrics are expected to generate tangible business outcomes. Specifically, WSU will improve its ability to market itself to new students, promoting a continuous cycle of improvement. This will ultimately raise WSU’s brand and position it for a bright future in a competitive market. Follow-on metrics include:

• Increased number of applications received for incoming freshman and transfer students
• Improved quality of the applicant pool (ability to attract students with higher GPA and SAT/ACT test scores)
• Improved quality of incoming freshman classes (better students will typically yield better student outcomes)

Technology

Specifically, for this client, we utilized the following technologies for data:

  • BM Cloud Pak for Data 3.5 running on Red Hat OpenShift 4.0 – to provide the underlying data layers with cloud platform flexibility.
  • IBM Cloud Pak for Data DataStage Cartridge for ETL
  • IBM API Connect – for data sharing as needed.
  • Prolifics Migration Factory Accelerator – This enables us to rapidly migrate ETL jobs from competitive platforms to IBM Cloud Pak for Data. This accelerator covers the gamut from analyzing the old systems and developing the new environment to testing, deployment and support. Prolifics can cost effectively provide this migration, cleansing and organization of data through our offshore component.
  • Prolifics Cloud Pak for Data OpenShift Accelerator – This speeds the interoperability between Cloud Pak for Data and Red Hat OpenShift. In collaboration with IBM Offering Management, we created this accelerator and documented the nuances to ensure a successful implementation of Cloud Pak for Data on OpenShift.

Specifically, for this client, we utilized the following technologies for AI:

  • IBM Watson Knowledge Catalog to prepare data for AI analysis.
  • Watson Assistant, Watson Discovery, and Watson Explorer to uncover, search and explore structured and unstructured client text and data across all channels.
  • Watson Studio and Watson Machine Learning to build and deploy automated AI workflows, and Auto AI to manage and simplify the AI lifecycle.
  • Prolifics Predictive Model Templates for Watson. This asset speeds the creation of custom predictive models by using a set of standard models as a starting point. Prolifics has created these standard models specifically for IBM Watson products.

NEXT STEPS

We’re fielding interest in this solution for a number of use cases across the whole spectrum of industries:

  • Retail – What indicators would gauge customer interest in a new product?
  • Healthcare – What would indicate a potential relapse and readmission of a recently discharged hospital patient?
  • Finance – What are the early warning signs that a customer may default on a loan?

About Prolifics

Prolifics is a global digital transformation leader with expertise in cloud, data and analytics, DevOps, digital business, and quality assurance across multiple industries. We provide consulting, engineering and managed services for all our practice areas at any point our clients need them – giving them fast, complete solution delivery experiences that they find nowhere else. Whether it’s initial advising and strategy; design and implementation; or ongoing analysis and guidance, Prolifics helps companies take charge of their digital future. Vision to Value. Faster. It’s not just the Prolifics’ tagline, it’s what drives us. Email solutions@prolifics.com or visit us at prolifics.com.


Share this article