KNN vs SVM: A Comparison of Algorithms

K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) are two common machine learning algorithms. Used for classifying images, the KNN and SVM each have strengths and weaknesses. When classifying an image, the SVM creates a hyper plane, dividing the input space between classes, classifying based upon which side of the hyperplane an unclassified object lands when placed in the input space. The KNN however, used as system of voting to determine which class an unclassified object belongs to, considering the class of the nearest neighbors in the input space. The SVM is extremely fast, classifying images in roughly ten seconds as opposed to the KNN which takes anywhere from forty to fifty seconds to classify the same image. When classifying the KNN generally does a good job, however, it generates a lot of small misclassifications that interfere with final classified image that is outputted as opposed to the SVM that will occasionally misclassify a large object that rarely interferes with the final classified image. While both algorithms yield positive results regarding the accuracy in which they classify the images, the KNN appears to be slightly more accurate