EM-ACO: Combining EM and ACO for Classification

Classification is a data mining technique that assigns data in a collection to target classes. Ant Colony Optimization (ACO) and Expectation Maximization (EM) algorithms are effective methods of machine learning which have been used widely for classification. In this work, a hybrid method is proposed by using these algorithms, as named EM-ACO. In the approach, ACO is first used to improve training data, and then EM classifies improved dada. We test our method on UCI machine learning data sets, iris and glass. The experimental results show proposed method performs better than EM

Keywords:

Classification , Hybrid method , Ant colony optimization algorithm , Expectation maximization algorithm