Double Ensemble kNN: Towards an Enhanced k-Nearest-Neighbors

Predictive Analytics, a branch of Data Science, has seen a surge in interest, attributed to the rise of Big Data and corporations? needs to identify consumer trends. Unsupervised Learning refers to a branch of Machine Learning in which unlabeled data is sorted into clusters, to identify trends and create target-segments. This can be used within Predictive Analytics to predict spam e-mails, which customers are likely to return, and more. Several unsupervised learning methods have been created, with K-means, an iterative clustering algorithm, being widely used due to its simplicity and stable nature. This paper creates three additional Ensemble methods using the K-Nearest-Neighbor as the base-learner. The Double-Ensemble, a novel method introduced here uses an Ensemble of kNNs with different ?k?s on subsets of data with randomized features. Experimental results demonstrated that the Double Ensemble method outperforms the normal kNN, with an additional accuracy of 9.8%.