Tool Image Searching Based Canopy K Means Clustering Using Web Programming Study Case MNC Publishing

Abstract

This research builds a tool for e-commerce search based on image input searching where the study case in this research is in MNC Publishing. The purpose of this research is to help users find the book that they want by inputting the image in the tool and then finding the book by the image that they find, and the result is linked to the detail of the book that they find. This research uses Canopy K Means Clustering to cluster the image and count the value of centers in system machine learning in this application to use clustering to minimize sum error compared by the count of centers value result by counting average images each cluster with reduced data images with value error loose distance each cluster until N iteration. Images normalization with reduced matrix from size 242193 reduce to 100 dimensions of matrix each data in web programming with 500 images result matrix input [500x100] dimensions system counts the image searching from size matrix [500x242193] become [500x100] dimensions of matrix for high computation with Canopy K Means Clustering. Testing with grayscale, the application got the result clustering 100% true for clustering from 50 data images and 77,4% from 500 data images.