Association Rule Mining is a Data Mining technique which extracts association rules from the given dataset. A good number of research work has been reported in Association Rule Mining algorithms which discovers positive association rules. Amongst them, only a few algorithms have focused on Association Rule Mining algorithms which discovers negative association rules too. Amongst the negative Association Rule Mining algorithms, most of them scans the dataset more than once to identify the frequent item sets and also doesn't guarantee that all the extracted rules are interesting. In order to overcome the above said challenges, EO-ARM, an Efficient and Optimized Positive-Negative Association Rule Mining algorithm has been proposed in this paper. EO-ARM produces both positive as well as negative association rules. It scans the dataset only once (irrespective of the size of dataset) to identify frequent item sets using a two dimensional matrix thereby increasing the efficiency. The two dimensional matrix is conceptually similar to k-map. It also optimizes the association rules by introducing a contingency matrix based correlation measure which prunes less interesting rules thereby overcoming the existing limitations. EO-ARM has been implemented using a binary transaction dataset. Several experiments were performed and an optimal support and confidence threshold has been identified for the given dataset. These optimized support and confidence threshold have been used to find the frequent item sets and generating rules from the dataset. Experimental results also proved that EO-ARM is more efficient in terms of execution time than the standard Apriori algorithm and more optimized in terms of no. of rules generated with the pruning done with the projected correlation measure. © 2014 IEEE.