Any efficient Bot-Detection tool must be able toclassify bot activity as, 'bot' with utmost accuracy. The key factorthat influences the efficiency of a Bot-Detection tool is theselection of a classification algorithm whose prediction accuracyis the maximum. This paper proposes the implementation of aHybrid Approach involving k Nearest Neighbor (kNN), NaïveBayes and ID3 classifiers resulting in most encouragingprediction accuracy values. The proposed scheme is followedafter the preprocessing phase that involves the computation ofweights of each attribute in the data set through different WeightBased Feature Selection algorithms and determining the mean ofweights of each attribute. A final Mean of mean values of all theattributes is computed. All those attributes whose mean valuesfall below the final Mean of mean values are discarded while allother remaining attributes are considered significant attributesand are used for botnet classification.