Opinion mining, a sub-discipline of information retrieval and concerns not with what a document is about, but with its expressed opinion. Feature selection is an important step in opinion mining, as customers express product opinions separately according to individual features. Statistical techniques like Correlation based Feature Selection (CFS) have been extensively used for feature selection to reduce the corpus size. Feature selection is NP hard and selecting features using statistical techniques is suboptimal. In this work a novel feature selection technique using Multi Objective Artificial Bee Colony algorithm is proposed. The proposed technique is evaluated using Fuzzy Unordered Rule Induction Algorithm (FURIA) classifier.  Results show improved classification accuracy in the proposed technique.