Making Explainable Decision Trees Faster & More Accessible
Balancing model accuracy with interpretability remains a central challenge in machine learning, particularly for classification tasks involving large datasets and continuous variables. This work examined how optimal sparse decision tree frameworks can be made more efficient and accessible without sacrificing their explainability or theoretical guarantees.
The research was presented at the Graduate Poster Exhibition during the 2025 SPARK! (Showcase of Projects, Art, Research, and Knowledge), a reimagining of Research Week that highlighted graduate research across disciplines. Developed within the Computer Science M.S. program at Rutgers University–Camden, the project was completed by Neha Mohan Kumar. The abstract below introduces her work on optimizing GOSDT-Guesses through a faster, memory-efficient Python implementation.
Abstract: Optimizing GOSDT-Guesses: A Faster, Memory-Efficient Python Implementation with LightGBM-Based Threshold Guessing
We are interested in studying machine learning frameworks for classification that are easily explainable. One such model, decision trees (and related tree models such as Random Forests), has been the basis of recent work by Dr. Cynthia Rudin’s research group at Duke University on Optimal Sparse Decision Trees (OSDT) for datasets with continuous variables. In a series of three papers beginning with OSDT, the group proposed algorithms to construct Generalized and Scalable Optimal Sparse Decision Trees (GOSDT) and GOSDT-Guesses, with the latter significantly improving performance while ensuring provable optimality with respect to model complexity and classification accuracy.
However, GOSDT models remain computationally expensive for large datasets and have limited accessibility due to their C++ implementation. In this work, an optimized version of GOSDT-Guesses was proposed that reduces time complexity and memory requirements while preserving accuracy. This was achieved through two key changes to the process of binarizing continuous variables, a prerequisite for constructing OSDTs.
First, the LightGBM framework was used instead of a traditional Gradient Boosting Machine (GBM) to guess thresholds. LightGBM employs techniques such as Gradient-based One-Side Sampling (GOSS), histogram-based learning, exclusive feature bundling (EFB), and a leaf-wise growth strategy to reduce computation time and memory usage. Second, the threshold selection process was refined through adaptive threshold grouping, which clusters similar values and retains representative thresholds, reducing the search space without compromising accuracy.
The work also included a complete reimplementation of GOSDT-Guesses in Python, eliminating the dependency on C++ bindings and making the algorithm more accessible to Python developers. The resulting implementation was significantly more memory-efficient and enabled the GOSDT approach to scale to larger datasets with continuous features. With reduced memory constraints, the number of estimators in LightGBM could be increased, further improving classification accuracy.
Graduate Poster Exhibition at SPARK!
The Graduate Poster Exhibition celebrates the research and creative work of the graduate community, showcasing everything from prose and code to original research and artistic expression. As part of SPARK! (Showcase of Projects, Art, Research, and Knowledge), a reimagining of Research Week, the exhibition highlights the depth, range, and impact of graduate scholarship and invites the campus community to engage with ideas taking shape across disciplines.
Interdisciplinary Insights: Rutgers-Camden’s Childhood Studies Programs
Rutgers University–Camden offers interdisciplinary M.A. and Ph.D. programs in Childhood Studies, focusing on the multifaceted experiences of children across various contexts. The M.A. program equips students with the skills to conduct research, influence social policy, and work with diverse child populations, leading to careers in public policy, social services, and education. The Ph.D. program immerses students in comprehensive theoretical and methodological training, preparing them for scholarly research and roles in academia, policy-making, and organizations dedicated to children’s welfare.
Crafting Stories, Analyzing Media: Explore the MA in English and Media Studies
Dive into the dynamic world of the Master of Arts in English and Media Studies at Rutgers University–Camden, where literature, culture, and digital media converge. This interdisciplinary program equips students with the skills to critically analyze texts, media, and cultural narratives while fostering creativity and research expertise. With courses spanning literary traditions, media analysis, and a unique focus on diversity, including topics like race and gender, the program prepares graduates for impactful careers in academia, publishing, and beyond. Plus, the vibrant Rutgers–Camden Writers House offers an inspiring space for creative growth through workshops, readings, and events, making this program a hub for innovative thinkers and storytellers.
Spotlights & Stories
Want more stories like this? Spotlights and Stories highlights alumni, faculty, staff, and student experiences through stories, video, and voice. Learn more
Explore Graduate School Programs
Rutgers Graduate School-Camden offers 20+ graduate certificates, master’s, and doctoral programs across various disciplines, including biology, data science, creative writing, and psychology. We take pride in our academic diversity. Learn more
Former Students, Get Involved
Stay connected with Rutgers-Camden! Engage with former students, attend events, and support current students. Your involvement strengthens our community and helps shape future success. Learn more