SPATIAL BIAS CORRECTION BASED ON GAUSSIAN KERNEL FUZZY C MEANS IN CLUSTERING

摘要: Clustering is the process of grouping data objects into set disjointed classes called clusters so that within a class are highly similar to one another and dissimilar in other classes. K-means (KM) Fuzzy c-means (FCM) algorithms popular powerful methods for cluster analysis. However, KM FCM have considerable trouble noisy environment inaccurate with large numbers different sample sized clusters. The Kernel based C-Means (KFCM) clustering moreover studied associated validity measures. Many numerical simulations used evaluate whether or not kernelized measures adequate ordinary ball-shaped Finally, new kernel functions Gaussian (GKFCM) proposed this research. GKFCM algorithm becomes generalized type of, BCFCM, KFCM presents more efficiency robustness. analysis scientific method utilizing unsupervised learning criteria segment given groups individuals. image segmentation its simplicity easiness implement. commonly hard c means fuzzy algorithm. It an important task areas such as machine learning, pattern recognition artificial intelligence. K well known widely partitioned easy implement, very efficient has linear time complexity. partitions dataset clusters, where object belongs only cluster. crisp membership function causes drawback result sensitive initial centers may converge local optima. environments, when difference size each large. There two approaches mitigate impact problems improve adaptively robustness On hand, many researchers concentrated on extending distance FCM. objective modified compensating strength homogeneities by allowing pixel labeling be influenced labels instant neighborhood. This modification Bias-corrected (BCFCM) But BCFCM also lacks noise it takes execution. So, version introduced solve drawbacks BCFCM. affected their parameters while Hence, work kernel-based proposed. provides than algorithms. rest paper about related section 2, methodology 3, experimental results 4, conclusion 5.