Research On Modern Optimization Algorithm Based K-Means Clustering And Its Applications On Student Grade Mining

Modern optimization algorithms are recently a hot research topic and they are already used in many areas. Simulated annealing and genetic algorithms are two most frequently used modern optimization algorithms. Simulated annealing algorithm is based on probabilities. Its idea is from the process of heating solid objects. Genetic algorithms are based on the simulation to Darwin's theory of evolution. It solves optimization problems similar to gene process such as selection, crossover and mutation.Data mining is a hot research topic as well. Clustering is an important sub-topic of data mining. K-means is a major algorithm for clustering. It has many advantages because it is easy to implement and close to clustering criteria. However, there is one disadvantage of K-means:it heavily relies on the beginning state due to local optimum. Modern optimization algorithms can deal with local optimum problem, so they are able to improve K-means clustering.In this paper, we focus on the application of modern optimization algorithms on improving K-means clustering. Firstly, modern optimization algorithms and K-means clustering algorithm are both investigated and analysis. Second, we investigated the previous researches about the application of modern optimization algorithms on K-means clustering, and their advantages and disadvantages are summarized. Third, based on our investigations, we propose simulated annealing and genetic algorithms which can be effectively used in student grade mining. Finally, we use the real data about student grades as the test data set. The experimental results on those real data set have proved the effectiveness of our proposed method. Moreover, simulated annealing and genetic algorithm based clustering methods have their own advantages and disadvantages:simulated annealing is sample and can easily be generalized, but genetic algorithm has better results. Both simulated annealing and genetic algorithms out-performs the basic K-means clustering method.Student grade mining is very important for improving education qualities. However, there are a lot of factors that impacts a student's grade, which makes the task of student grade mining more difficult. With the help of simulated annealing or genetic algorithms, we can get more higher accurate clustering results. This might provide a useful reference to student developments.