K-Nearest Neighbor for Identification of Pulsars in Astronomy
In radio astronomy, pulsar detection is not an easy method because most radiation detectors have telescopic noise. One of the methods of cognitive demand is to decide on the proper radiation to be released by the pulsar. In this paper, the K-Nest Neighbor based classifier has been implemented to identify pulsar stars by noise by classifying pulsar candidates from non-pulsar candidates. The nearest-neighbor concept is explained with an assortment of data. The algorithm is implemented using four statistical values of two input features. The statistics are standard deviation, excess kurtosis and skewness. The input features are integrated profile and DM-SNR curve. 17,897 observations are used to train the classification, and the average accuracy obtained is 96.54%. A detailed explanation of the algorithm is also given.