Detecting outliers using KNN
The KNN algorithm is typically used in a supervised learning setting where prior results or outcomes (labels) are known.
It can be used to solve classification or regression problems. The idea is simple; for example, you can classify a new data point, Y, based on its nearest neighbors. For instance, if k=5, the algorithm will find the five nearest data points (neighbors) by distance to the point Y and determine its class based on the majority. If there are three blue and two red nearest neighbors, Y is classified as blue. The K in KNN is a parameter you can modify to find the optimal value.
In the case of outlier detection, the algorithm is used differently. Since we do not know the outliers (labels) in advance, KNN is used in an unsupervised learning manner. In this scenario, the algorithm finds the closest K nearest neighbors for every data point and measures the average distance. The points with the most significant distance from the population...
