Active Learning for kNN Using Instance Impact

Labelling unlabeled data is a time-consuming and expensive process. Labelling initiatives should select samples that are likely to enhance the classification accuracy of the classifier. Several methods can be employed to accomplish this goal. One of these techniques is to select samples with the highest level of uncertainty in their predicted labels. Experts then label these samples. Another option is to choose samples at random. This paper proposes three methods for identifying unlabeled samples to improve predictive accuracy when they are labelled. Our study explores how to select samples when we have very few labelled samples available from manifold distributed data sets. In order to assess performance, we have compared our approaches with uncertainty sampling and random sampling. We demonstrate that our methods outperform uncertainty sampling and random sampling by using public and real-world data sets.