Improved and Effective Artificial Bee Colony Clustering Algorithm for Social Media Data (I-ABC)

Social media data made real world like a web of data which is highly categorical in nature. Data having categorical attributes are omnipresent in existing real world. Clustering is an effective approach to deal with categorical data. However, partitional clustering algorithms are prone to fall into local optima for categorical data. A novel approach of ABC K-modes has been proposed to address this issue but acceleration issue of this algorithm was still a challenge for it. In this paper, we address this challenge to reduce the acceleration factor of algorithm and proposing a novel modified ABC K-modes approach which we refer as N-ABC Kmodes approach. In our approach, unlike existing ABC K-modes we introduces different attribute matrix for each data sets. In further step, we apply XOR operation to combine the matrix of similar attributes. In last phase, dissimilar data would form a cluster and we apply clustering follow by searching on this cluster. The performance of New ABC K-modes evaluated by a series of tests and experiments over real time streaming social media data like twitter and facebook in comparison with that of other popular algorithms for categorical data.