Active Learning for Decision Making

Abstract

This paper addresses focused information acquisition for predictive data mining. As businesses strive to cater to the preferences of individual consumers, they often employ predictive models to customize marketing efforts. Building accurate models requires information about consumer preferences that often is costly to acquire. Prior research has introduced many â active learningâ policies for identifying information that is particularly useful for model induction, the goal being to reduce the acquisition cost necessary to induce a model with a given accuracy. However, predictive models often are used as part of a decision-making process, and costly improvements in model accuracy do not always result in better decisions. This paper develops a new approach for active information acquisition that targets decision-making specifically. The method we introduce departs from the traditional error-reducing paradigm and places emphasis on acquisitions that are more likely to affect decision-making. Empirical evaluations with direct marketing data demonstrate that for a fixed information acquisition cost the method significantly improves the targeting decisions. The method is designed to be genericâ not based on a single model or induction algorithmâ and we show that it can be applied effectively to various predictive modeling techniques.