Exploiting Temporal Structure for Causal Modeling

The need for modeling causality, beyond mere statistical correlations, for meaningful application of data mining to real world problems has been recognized widely. The framework of Bayesian networks, along with the related causal networks, is well suited for the modeling of causal structure, and its applicability to various application domains has been well investigated. Many of these application areas naturally involve temporal data, and the question of how best to leverage the temporal information in the data for causal modeling naturally arises. In the present paper we propose a novel approach to exploiting temporal information in causal modeling. Our methodology is based on the intuition that a cause invariably precedes its effects, and uses this to help boost the performance of causal modeling. Specifically, we present a family of algorithms that exploit the aforementioned temporal constraint to orient the directionality of variable dependencies. We compare our approach to a group of related methods, and demonstrate consistent gains in modeling accuracy by conducting systematic experiments using synthetic data. We also apply the proposed methods to a real world data set involving key performance indicators of corporations, and present some concrete results.