Prof. Igor Kononenko, Faculty of Computer and Information Science, University of Ljubljana, Slovenia
Explaining individual predictions of an arbitrary model
After a brief overview of our research in the field of machine learning and data mining I will present a general method for explaining individual predictions of arbitrary classification and regression models. The method is based on fundamental concepts from coalitional game theory and predictions are explained with contributions of individual feature values. All subsets of input features are perturbed, so interactions and redundancies between features are taken into account. We overcome the method’s initial exponential time complexity with a sampling-based approximation. The method is efficient and its explanations are intuitive and useful. Besides individual predictions, the method can also be used to explain the whole models.
Igor Kononenko is a professor at the Faculty of Computer and Information Science, University of Ljubljana, Slovenia. He is the Head of Laboratory for Cognitive Modelling and the Head of Artificial Intelligence Department. His major research interest is machine learning and data mining. He is the (co)author of about 225 papers in these fields and 13 textbooks, including the book published by Horwood: Machine Learning and Data Mining - Introduction to Principles and Algorithms. Kononenko is a member of editorial boards of two journals and was a PC member of more than 30 scientific conferences.