Optimization of the weights of weighted naïve Bayesian classifier

The rapid development of the technology, along with the increasing amount of data, makes data analysis inconvenient. Today, it is important that many processes can be recorded, stored and accessed in an electronic environment. As long as the data are not processed, it does not make any sense. Data mining is used to make the data meaningful. Data mining is the process of retrieving useful information from large-scale data by separating the information among the large-scale data and retrieving the data by using a software to make predictions about the future. In this thesis, methods of data mining and Bayes classification algorithm are examined. The Weighted Bayes algorithm, which is an improved model of the Bayes classification algorithm, is also examined in the thesis and it is aimed to optimize the weights used in this method. For this purpose, we propose Fasted Weighted Naive Bayes (FW-NB) method to find weights quickly and then weights are optimized by using Genetic Algorithm (Genetic Algorithm Based Weighted Naive Bayes-GAW-NB) for the optimization of weights. The methods used in this thesis are applied on 5 different data sets and the results are evaluated comparatively. The results show that FW-NB algorithm is faster than W-NB algorithm and performance of GAW-NB algorithm is higher than W-NB. Key Words: Data Mining, Classification Algorithms, Bayes Algorithms, Weighted Naïve Bayes Algorithms