Towards group behavioral reason mining

Many studies have been proposed to research behavior mining. However, in many cases, the aim of exploring behaviors is to exploit their motivations. Based on discovered behavioral reasons, we are able to conduct subsequent actions to impel or impede those behaviors. Although some logical approaches have been proposed to derive an explanation for a set of observations using abductive reasoning, there are few methods that take a statistical approach for group behavioral reason mining. Statistical methods enable us to discover behavioral reasons automatically in an uncertain situation. To address this issue, we propose a computational model and a family of algorithms called BRMA (Behavioral Reason Mining Algorithm), which exploits various distance functions to discover group behavioral reasons in three statistical ways. The BRMA algorithms have low time complexity and run extremely fast. Based on two datasets, we conducted comprehensive experiments to evaluate the effectiveness of the BRMA algorithms. The empirical experimental results indicate that the BRMA algorithms have a relatively high accuracy, and that among the BRMA family, BRMAMP outperforms BRMAAverage and BRMAWeight.

► We develop a group behavior model and exploit various distance functions for group behavioral reason mining. ► A family of algorithms called BRMA is developed to discover the group behavioral reasons based on similarity. ► Based on two datasets, a comprehensive experiment is conducted to evaluate the efficiency and the efficacy of the BRMA algorithms. ► The BRMA algorithms have a relatively high accuracy and BRMAMP outperforms BRMAAverage and BRMAWeight in terms of accuracy. ► The time complexity of the BRMA algorithms is very low, making them extremely efficient.