Principle-Component-Analysis

In this project, I have tried to compare the performance of the KNN algorithm and then carried out PCA in detail. PCA is a ML Algorithm which carries the best possible lower dimension so the core essence of the data can be captured. In this notebook, the benefit we have gained using PCA is that the execution has been very quick and steady, as well as that 784 dimensional data is visualized in 2D and 3D. Furthermore, we dived deeper into mathematical components behind PCA such as EigenVectors and EigenValues. For example, for the 3D plot, the best (highest) lambda values were chosen. The dimension of our data was from reduced for better visualization. Moreover, accuracy scores were calculated using various n_components inside the PCA function, and they can be compared with the accuracy score of KNN. Also, unexplained variance concept was clearly defined inside the notebook. Finally, a graph was plotted which told us the percentage of variance being explained by the number of components. Cumulatively, 90% was to be reached for best explanation.