Online active learning of decision trees with evidential data

Abstract

Learning from uncertain data has been drawing increasing attention in recent years. In this paper, we propose a tree induction approach which can not only handle uncertain data, but also furthermore reduce epistemic uncertainty by querying the most valuable uncertain instances within the learning procedure. We extend classical decision trees to the framework of belief functions to deal with a variety of uncertainties in the data. In particular, we use entropy intervals extracted from the evidential likelihood to query selected uncertain querying training instances when needed, in order to improve the selection of the splitting attribute. Our experiments show the good performances of proposed active belief decision trees under different conditions.

An IoT-Based Intelligent Wound Monitoring System

“…Data pick up is itself judged by measure called entropy , which utilized to measures the impurity of set of training objects. For a collection of data set S, entropy formula is given in equation 1, formulated by Claude Shannon , also used by [28,[35][36][37][38]. Entropy helps decision tree to determine how informative a node is.…”

Parametric classification with soft labels using the evidential EM algorithm: linear discriminant analysis versus logistic regression

“…A second category of applications concerns the problem of learning from uncertain data. For instance, Sutton-Charani et al [44,45] and Ma et al [30] applied the E 2 M algorithm to decision tree inference from data with uncertain attributes, while the problem of clustering data with fuzzy attributes was considered by Quost and Denoeux in [38]. In [48], the authors considered the modeling of lifetime data using mixture models and progressively censored observations.…”

“…Such a procedure, called the Evidential EM (E 2 M) algorithm, was introduced in [12]; it extends the Expectation-Maximization (EM) algorithm [10] by allowing for maximum-likelihood estimation from partially missing data. The E 2 M algorithm has been applied to a variety of tasks and models, including partially supervised Independent Factor Analysis [5], Hidden Markov Models with partially hidden states [39], fuzzy data clustering using Gaussian mixture models [38], decision trees [44,45,30], and mixture models with progressively censored data [48].…”

Belief Function Definition for Ensemble Methods - Application to Pedestrian Detection in Dense Crowds

“…Recent work highlighted that intrinsic uncertainty related to learning as well as uncertainty due to imprecise data may be jointly managed inside the decision tree by defining entropy intervals from evidential likelihood [14].…”