Disjunctive combined causal rules mining

Causal discovery is a well-studied problem due to an urgent need for systems that predict, explain, and make proper and necessary decisions in many domains including epidemiology, biology, medicine, economics, physics, and social sciences. Existing techniques such as learning Bayesian networks (BNs) and Randomized controlled trials (RCTs) are expensive and time consuming. In addition, they only find single cause rules from certain data. There are numerous important applications wherein we have to generate disjunctive causal rules from uncertain data. In this paper we propose an algorithm called DCCRUD that employs frequent itemsets mining algorithms to discover disjunctive combined causal rules from uncertain data. To the best of our knowledge ours is the first paper to address this important problem. Discovering causal rules where targets are disjunctions of variables might be equally important. DCCRUD applies to uncertain databases. We evaluate the performance of the proposed algorithms on real datasets.