The recent rise of machine learning (ML) has been leveraged by practitioners and researchers to provide new solutions to an ever growing number of business problems. As with other ML applications, these solutions rely on model selection, which is typically achieved by evaluating certain metrics on models separately and selecting the model whose evaluations (i.e., accuracy-related loss and/or certain interpretability measures) are optimal. However, empirical evidence suggests that, in practice, multiple models often attain competitive results. Therefore, while models’ overall performance could be similar, they could operate quite differently. This results in an implicit tradeoff in models’ performance throughout the feature space which resolving requires new model selection tools. This paper explores methods for comparing predictive models in an interpretable manner to uncover the tradeoff and help resolve it. To this end, we propose various methods that synthesize ideas from supervised learning, unsupervised learning, dimensionality reduction, and visualization to demonstrate how they can be used to inform model developers about the model selection process. Using various datasets and a simple Python interface, we demonstrate how practitioners and researchers could benefit from applying these approaches to better understand the broader impact of their model selection choices.

We develop a method for understanding specific predictions made by (global) predictive models by constructing (local) models tailored to each specific observation (these are also called “explanations" in the literature). Unlike existing work that “explains” specific observations by approximating global models in the vicinity of these observations, we fit models that are globally-consistent with predictions made by the global model on past data. We focus on rule-based models (also known as association rules or conjunctions of predicates), which are interpretable and widely used in practice. We design multiple algorithms to extract such rules from discrete and continuous datasets, and study their theoretical properties. Finally, we apply these algorithms to multiple credit-risk models trained on the Explainable Machine Learning Challenge data from FICO and demonstrate that our approach effectively produces sparse summary-explanations of these models in seconds. Our approach is model-agnostic (that is, can be used to explain any predictive model), and solves a minimum set cover problem to construct its summaries.

Dimension reduction (DR) techniques such as t-SNE, UMAP, and TriMap have demonstrated impressive visualization performance on many real-world datasets. One tension that has always faced these methods is the trade-off between preservation of global structure and preservation of local structure: these methods can either handle one or the other, but not both. In this work, our main goal is to understand what aspects of DR methods are important for preserving both local and global structure: it is difficult to design a better method without a true understanding of the choices we make in our algorithms and their empirical impact on the low-dimensional embeddings they produce. Towards the goal of local structure preservation, we provide several useful design principles for DR loss functions based on our new understanding of the mechanisms behind successful DR methods. Towards the goal of global structure preservation, our analysis illuminates that the choice of which components to preserve is important. We leverage these insights to design a new algorithm for DR, called Pairwise Controlled Manifold Approximation Projection (PaCMAP), which preserves both local and global structure. Our work provides several unexpected insights into what design choices both to make and avoid when constructing DR algorithms.