Multi-objective optimization of document similarity results with(via) nsga-ii

Increasing performance of classification algorithms is one of the important goals of data mining. In this thesis study, it has been investigated how to increase the data mining classification success with heuristic methods. The training data set used in the classification is optimized in terms of the both similarity calculation results and the classification ability. It is aimed to maximize the similarity results of the vectors in the same class and at the same time to minimize the similarity results of the vectors in different classes. For these two contradictory cases, Non Sorting Genetic Algorithm II (NSGA II), which is a multipurpose heuristic method, is used. It is aimed to approximate the false classification ratios zero in each iteration of the optimization. In this study, care was taken in order to perform all phases of data mining in order. The raw data were processed and attributes were extracted. For size reduction operations, Principal Component Analysis (PCA) is used. Using the K Nearest Neighborhood (KNN) Algorithm on the data sets, the lean classification successes and the post-optimization classification successes are compared. Optimization has been shown to increase the ability to classify training data sets. An increase in classification success was observed when the optimized data were used as a training set.