CAREER: Novel Data Analytics and Optimization Approaches for Improving Efficiency of Mental Health Resources

About This Grant

This Faculty Early Career Development Program (CAREER) grant will contribute to scientific progress and to the advancement of national prosperity and welfare by supporting research on efficient scalable methods for characterizing and optimizing large-scale non-stationary stochastic systems, as well as modeling and predicting stochastic processes with time-dependent inputs. These problems arise in a wide array of critical applications, including environmental surveillance, personalized healthcare, and manufacturing process control. Under this award, the target application focuses on ensuring effective and efficient use of mental health resources on university campuses. This award pursues a fundamental understanding of stochastic systems that can be more effectively characterized through a novel analytical–computational approach. The accompanying educational plan aims to improve the academic success of students with challenges and to increase the engagement of students in STEM education through research integration, mentorship, and dedicated fast-track programs. This research seeks to establish an innovative Fourier-based framework to accurately characterize the transient behavior of large-scale non-stationary stochastic systems. The methodology leverages Euler’s formula to efficiently describe system behavior through a condensed form of the Chapman–Kolmogorov forward equations. A major advantage of this approach is that it incorporates crucial gradient information, which will be embedded in a gradient-based method, thereby enabling efficient optimization of complex stochastic systems. Additionally, this research looks to introduce a novel class of Gaussian process regression techniques that integrate time-varying covariates, facilitating the analysis of complex, temporally correlated data to predict student outcomes under different treatments. This effort will be complemented by a Bayesian optimization-based active learning framework aimed at constructing maximally effective treatments through a novel epsilon-optimal spatial branching scheme grounded in a new class of efficient convex underestimators. Together, these advancements intend to be used to expand students’ access to critical mental health services and to design treatment plans tailored to each student’s specific needs. The performance assessments of the project activities will be informed by data obtained through collaborations with a large-scale network of college counseling centers. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.