Problem Statement

More than 30 percent of the first-year students do not return.

Solution

Identify and help at-risk students at the beginning of the semester. In addition, admit students who are more likely to succeed. AI makes this possible.

The University can assess the likelihood of student success and, therefore, retention before the students arrive at the campus using correlations between their pre-college characteristics and retention predictors. For example, the correlation between high school GPA and the registration variables is statistically significant. The data show that students who take fewer than 15 credits and wait longer to register are less likely to return. In addition to high school GPA, an essay in the application packet may also reflect their pre-college characteristics.

Student Retention Predictors

Takeaways

Return predictors on the left, while dropout predictors on the right.

The higher the high school GPA a student has, the more likely the student will return.

Students with a problem in the dorm are less likely to return.

Students who register 14 credits or fewer are less likely to return.

The longer a student delays registration, the less likely the student will return.

Students who applied late for admission (PIDM_KEY > 327631) are less likely to return, while students who applied early (PIDM_KEY < 323066) are more likely to return.

• PIDM_KEY Students are assigned to their ID in the order they applied. Student A who applied to PSU for admission after Student B would have a higher number than Student B. This finding has not yet been confirmed by application date data and will be examined in the future.

Pre-College Characteristics and Retention Predictors

Pre-College Characteristics and Retention Predictors

In top 10 high risk students, the model shows 80% accuracy.

Student 1 is predicted to drop out with a 73% probability. Supporting this prediction are factors including registration of 14-credits-or-fewer credits and registration delay of more-than-59-days. Contradicting it are factors including no dorm room change.

Student 2 is predicted to drop out with a 54% probability. Supporting this prediction are variables including registration of 14-credits-or-fewer credits and registration delay of more-than-59-days. Contradicting it are no dorm room change and the residency (i.e., the student is an NH resident).

Plot

Takeaways

The biggest saving opportntity lies in nonresident undeclared freshmen.

The cost calculation uses the current tuition rates of AY 2022: resident tuition of $14,492, non-resident tuition of $24,432, and room and board charges of $11,580. It assumes that all students pay the full tuition.