While positive association rules identify the co-occurrences of frequent item-sets, negative association rules find out the negation relationships of frequent items-sets by forming occurrence of an item-set characterized by the absence of others. The notion of negative association rule is very useful in customer-driven domains such as market basket analysis for identifying products that conflict with each other. When data are scattered in multiple data sources that are located in different regions, negative relationships among frequent item-sets are very important for arriving decisions both at strategic and branch levels. This paper presents an approach for mining negative association rules from multiple data sources and synthesizing global negative association rules voted by most of the data sources. The proposal has been justified experimentally by using a bench marked database found in UCI machine learning repository. © 2013 Springer.