A group decision making-based parallel multi-label feature selection for big data applications

Extracting knowledge from data is challenging in real-world scenarios. Particularly in big data learning, the volume of the data is a significant challenge. Therefore, scalable feature selection algorithms are essential for datasets of this scale. In this paper, a parallelized multi-label feature selection algorithm is proposed using intuitionistic fuzzy sets (IFSs) and group decision making. This partitions data instances into multiple groups. Then the mutual information between each feature and label within each group is computed. Subsequently, the mutual information values are transformed into IFSs and aggregated using the intuitionistic fuzzy weighted averaging (IFWA) operator. Parallel calculations were employed for this, which makes it suitable for large number of instances. Finally, the rank of the features is obtained using the intuitionistic fuzzy Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method. This opens a new perspective on the feature selection by adapting group decision making for parallel execution. The experiments conducted on 15 datasets show the effectiveness of the method.