Research On Fuzzy C-means Clustering Based On Differential Evolution

With the rapid development of database and information technology, the amount of data that we faced has grown explosively. In order to make better use of resources and obtain valuable information from these data, data mining technology is introduced. Currently, clustering analysis is a very important technology in the field of data mining. Many effective clustering algorithms have been proposed. Among them, the fuzzy C-Means(FCM) algorithm is relatively well and it is widely used. However, this algorithm highly depends on the initial value and also is sensitive to noise data. Differential evolution(DE) algorithm is a simulation of biological evolution random search algorithm. Its advantages lie in simple process, few control parameters, easy to implement, strong global convergence and robustness. Differential Evolution based fuzzy C- means clustering(FCDE) algorithm is applied the DE algorithm to FCM algorithm. It addresses the issue of depending on the initial value and sensitive to noise data in a certain degree. However, due to the randomness of mutation and crossover operations, the convergence of FCDE algorithm is very slow. Therefore, in view of the above problems, this thesis deeply investigates FCDE algorithms and the main its in-depth research. The main contents are as follows:1. In view of the issue of large granularity of crossover operation in DE algorithm, this thesis introduces a new crossover operator and change the crossover operation from the original sample data to the dimensions of the sample data, improving the precision of solutions; In view of the sensibility of factors F and CR of crossover operation in DE algorithm, this thesis introduces an adaptive updating operation, which makes the F and CR dynamically updating during the execution of the algorithm, instead of the original constant. This debases the influence of the sensitivity of F and CR on DE algorithm.2. In view of the issue of slow convergence of DE algorithm, this thesis introduces a new mutation operator and uses a greedy strategy, making the DE algorithm fast converging to the optimal solution. In order to prevent it from falling into local optimum, this thesis introduces a mutation mechanism and makes the DE algorithm eventually converging to the global optimum. Finally, the improved DE algorithm is applied to the clustering problem and an improved differential evolution based on fuzzy C- means clustering algorithm-1(IFCDE-1) is proposed in this thesis.3. Also, this thesis proposes another algorithm to address the issue of slow convergence of DE algorithm. We combine the FCDE algorithm and FCM algorithm, utilize the advantages of both algorithms(FCDE algorithm does not depend on the initial value and convergence of FCM algorithm is fast) and propose another improvement differential evolution based fuzzy C- means clustering algorithm-2(IFCDE-2).4. In order to verify the effectiveness of the proposed algorithms, we use the three real data sets, which are open source, and two artificial data sets to conduct the simulation. The results show that the proposed algorithms can improve the convergence efficiency and also obtain accuracy solutions.