Random Forest Similarity Maps: A scalable visual representation for global and local interpretation

Abstract

Machine Learning prediction algorithms have made significant contributions in today’s world, leading to increased usage in a variety of domains. However, as ML algorithms surge, the need for transparent and interpretable models becomes essential. Visualizations have shown to be instrumental in increasing model transparency, allowing users to grasp models’ inner workings. Despite their popularity, visualization techniques still present visual scalability limitations, mainly when applied to analyze complex models, such as Random Forests (RF). In this work, we propose Random Forest Similarity Map (RFMap), a scalable visual analytics tool designed to analyze RF models. RFMap focuses on explaining the inner working mechanism to users in a simplistic way through three views, describing data instances’ predictions, presenting an overview of the forest of trees, and highlighting instances’ feature values. The interactive nature of RFMap allows users to visually interpret models’ errors and decisions, establishing the confidence and users’ trust in RF models.