2015/06/08 by Viktoriya Krakovna, Krakovna, Viktoriya, Jiong Du +3
Computer Science · #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Machine Learning (stat.ML) #Machine Learning and Data Classification #Time Series Analysis and Forecasting
paper · pdf · doi:10.48550/arxiv.1506.02371
openalex publication_date 2015/06/08 · openalex created_date 2022/10/03 · openalex updated_date 2026/07/28
It is becoming increasingly important for machine learning methods to make\npredictions that are interpretable as well as accurate. In many practical\napplications, it is of interest which features and feature interactions are\nrelevant to the prediction task. We present a novel method, Selective Bayesian\nForest Classifier, that strikes a balance between predictive power and\ninterpretability by simultaneously performing classification, feature\nselection, feature interaction detection and visualization. It builds\nparsimonious yet flexible models using tree-structured Bayesian networks, and\nsamples an ensemble of such models using Markov chain Monte Carlo. We build in\nfeature selection by dividing the trees into two groups according to their\nrelevance to the outcome of interest. Our method performs competitively on\nclassification and feature selection benchmarks in low and high dimensions, and\nincludes a visualization tool that provides insight into relevant features and\ninteractions.\n