Comparison of Fuzzy C-Means, Fuzzy Kernel C-Means, and Fuzzy Kernel Robust C-Means to Classify Thalassemia Data

Abstract

Among the inherited blood disorders in Southeast Asia, thalassemia is the most prevalent. Thalassemias are pathologies that derive from genetic defects of the globin genes. Thalassemia is also considered a health burden among the world's population. Thalassemia cannot be cured, but there is a method to prevent the occurrence of thalassemia by early detection with screening. The aim is to identify the suspected unrecognised diseases in a population that seems healthy and asymptomatic using tests, examinations, or other procedures that can be applied quickly and easily to the target population. Research on thalassemia has been done extensively, such as testing the accuracy of β-thalassemia data in Thailand using the Bayesian Network and Multinomial Logistic Regression. In this study, we will compare the performance of the classification of thalassemia data by Fuzzy C-Means, Fuzzy Kernel C-Means, and Fuzzy Kernel Robust C-Means. The author uses thalassemia data from Indonesia, acquired from Harapan Kita Children and Womens's Hospital, Jakarta, that consists of 82 thalassemia samples from the patients of thalassemia and 68 nonthalassemia samples with 11 features. In total, there are 150 data patients used in this paper. The results show the accuracy of the classification. The accuracy of FCM is 100% when training data is 90%, FRCM is 100% when training data is 90%, and FKRCM, which is the modified Fuzzy, 100% when we use the = 0.0001 and 80% & 90% training data. This result denote that Fuzzy C-Means, Fuzzy Robust C-Means, and Fuzzy Kernel Robust C-Means perfectly classify thalassemia data from Indonesia.

Lung cancer classification using fuzzy c-means and fuzzy kernel C-Means based on CT scan image

“…A data set is transformed into a new feature space using a kernel with a higher space [24]. Therefore, the non-linear problem generalized, in combination with linear models, is overcome [13]. Let ∈ is the original data set.…”

“…For a data set = { 1 , 2 , … , } ⊆ , × membership matrix = [ ], 1 ≤ ≤ , 1 ≤ ≤ and cluster center = { 1 , 2 , … , } where every object in V is a part of ddimensional Euclidean Space [24]. Their objective functions are as [13], [14]:…”

“…In addition, previous research on the classification of lung nodules was performed with various methods such as convolutional neural network [8]- [10], support vector machine [11], and semi-supervised adversarial model [12]. The fuzzy C-Means method was initially used to classify thalassemia data [13], breast cancer [14], and intrusion detection system [15], while the fuzzy kernel C-Means was for chronic sinusitis [16], insolvency prediction [17], direction and indonesian stock price movement [18].…”

Comparison some of kernel functions with support vector machines classifier for thalassemia dataset

“…There are 7% of the world's population as carriers of thalassemia with the death of about 50,000-100,000 children [3]. In Indonesia, thalassemia is one of the most common chronic diseases [4]. Currently, thalassemia ranks 5th among non-communicable diseases after heart disease, cancer, kidney, and stroke with the number of carriers 3.8% of the entire population in Indonesia.…”

“…There are some methods on previous researches to classify thalassemia, such as fuzzy kernel robust C-means, fuzzy C-means, and fuzzy kernel C-means [4], neural networks and genetic programming [10], artificial intelligence algorithms [11], artificial neural network [12], and naïve bayes [13]. Also, [12], [14] used SVM that showed good result with 93.2% accuracy and 100% AUC respectively.…”