Mass estimation and its applications

This paper introduces mass estimation--a base modelling mechanism in data mining. It provides the theoretical basis of mass and an efficient method to estimate mass. We show that it solves problems very effectively in tasks such as information retrieval, regression and anomaly detection. The models, which use mass in these three tasks, perform at least as good as and often better than a total of eight state-of-the-art methods in terms of task-specific performance measures. In addition, mass estimation has constant time and space complexities.