Problems Identification: This project involves the implementation of efficient and effective KNN classifiers on MNIST data set. The MNIST data comprises of digital images of several digits ranging from 0 to 9. Each image is 28 x 28 pixels. Thus, the data set has 10 levels of classes.

Generally, transformation with 59 component(85% variance explained) is efficient in terms of low error rate and low query time, it is by far the best in this study. Accuracy rate is 98%, query time is 29 milliseconds.