Joon S. Lim

Letters: A neuro-fuzzy approach for diagnosis of antibody deficiency syndrome

This paper presents a neuro-fuzzy approach for diagnosis of antibody deficiency syndrome, where a new neuro-fuzzy network with fuzzy activation functions (FAFs) at hidden layer is used. The FAFs capturing some essential information on pattern distributions, can be adaptively constructed using training examples. To improve the generalization capability and reduce the model complexity, a heuristic method for feature selection is proposed by measuring the size of non-overlapped areas of the FAFs. The effectiveness of our proposed techniques is investigated by an immunology clinical data set collected from the University of California, Irvine (UCI) immunology laboratory.

Classification of normal and epileptic seizure EEG signals using wavelet transform, phase-space reconstruction, and Euclidean distance

This paper proposes new combined methods to classify normal and epileptic seizure EEG signals using wavelet transform (WT), phase-space reconstruction (PSR), and Euclidean distance (ED) based on a neural network with weighted fuzzy membership functions (NEWFM). WT, PSR, ED, and statistical methods that include frequency distributions and variation, were implemented to extract 24 initial features to use as inputs. Of the 24 initial features, 4 minimum features with the highest accuracy were selected using a non-overlap area distribution measurement method supported by the NEWFM. These 4 minimum features were used as inputs for the NEWFM and this resulted in performance sensitivity, specificity, and accuracy of 96.33%, 100%, and 98.17%, respectively. In addition, the area under Receiver Operating Characteristic (ROC) curve was used to measure the performances of NEWFM both without and with feature selections.

Detecting ventricular arrhythmias by NEWFM

The ventricular arrhythmias including ventricular tachycardia (VT) and ventricular fibrillation (VF) are life-threatening heart diseases. This paper presents an approach to detect normal sinus rhythm (NSR) and VF/VT using the neural network with weighted fuzzy membership functions (NEWFM). NEWFM classifies NSR and VF/VT beats by the trained bounded sum of weighted fuzzy membership functions (BSWFMs) using one input features from the Creighton University Ventricular Tachyarrhythmia Data Base (CUDB). In this paper, six input features are obtained from two steps. In the first step, 8s original ECG signal are transformed by Haar wavelet function, and then 256 coefficients of d3 at levels 3 are obtained. In the second step, six input features are obtained by phase space reconstruction (PSR) algorithm using 256 coefficients of d3 at levels 3. The one generalized feature is extracted by the non-overlap area distribution measurement method. The one generalized feature is used for the VF/VT data sets with reliable accuracy and specificity rates of 90.1% and 92.2%, respectively.

A 2-D Visual Model for Sasang Constitution Classification Based on a Fuzzy Neural Network

The human constitution can be classified into four possible constitutions according to an individual’s temperament and nature: Tae-Yang (太陽), So-Yang (少陽), Tae-Eum (太陰), and So-Eum (少陰). This classification is known as the Sasang constitution. In this study, we classified the four types of Sasang constitutions by measuring twelve sets of meridian energy signals with a Ryodoraku device (良導絡). We then developed a Sasang constitution classification method based on a fuzzy neural network (FNN) and a two-dimensional (2-D) visual model. We obtained meridian energy signals from 35 subjects for the So-Yang, Tae-Eum, and So-Eum constitutions. A FNN was used to obtain defuzzification values for the 2-D visual model, which was then applied to the classification of these three Sasang constitutions. Finally, we achieved a Sasang constitution recognition rate of 89.4 %.

Economic turning point forecasting using neural network with weighted fuzzy membership functions

This paper proposes a new forecasting model based on neural network with weighted fuzzy membership functions (NEWFM) concerning forecasting of turning points in business cycle by the composite index. NEWFM is a new model of neural networks to improve forecasting accuracy by using self adaptive weighted fuzzy membership functions. The locations and weights of the membership functions are adaptively trained, and then the fuzzy membership functions are combined by bounded sum. The implementation of the NEWFM demonstrates an excellent capability in the field of business cycle analysis.

Discrimination of ventricular arrhythmias using NEWFM

The ventricular arrhythmias including ventricular tachycardia (VT) and ventricular fibrillation (VF) are life-threatening heart diseases. This paper presents a novel method for detecting normal sinus rhythm (NSR), VF, and VT from the MIT/BIH Malignant Ventricular Arrhythmia Database using the neural network with weighted fuzzy membership functions (NEWFM). This paper separates pre-processing into 2 steps. In the first step, ECG beasts are transformed by using Filtering Function [1]. In the second step, transformed ECG beasts produce 240 numbers of probability density curves and 100 points in each probability density curve using the probability density function (PDF) processing. By using three statistical methods, 19 features can be generated from these 100 points of probability density curve, which are the input data of NEWFM. The 15 generalized features from 19 PDF features are selected by non-overlap area measurement method [4]. The BSWFMs of the 15 features trained by NEWFM are shown visually. Since each BSWFM combines multiple weighted fuzzy membership functions into one using bounded sum, the 15 small-sized BSWFMs can realize NSR, VF, and VT detection in mobile environment. The accuracy rates of NSR, VF, and VT is 98.75%, 76.25%, and 63.75%, respectively.

Bhattacharyya Distance for Identifying Differentially Expressed Genes in Colon Gene Experiments

Identify a small number of differentially expressed genes for accurate classification of gene samples is essential for the development of diagnostic tests. We present an approach for cancer molecular feature selection method based on their gene expression profiles. Tumor and normal colon tissues were classified in this research. The Bhattacharyya distance was used as the gene selection method to identify the small number of differentially expressed genes for the colon cancer analysis. Finally we selected 7 genes for the colon cancer analysis with 95.16% accuracy by using a fuzzy neural networks classifier. Compare with other colon cancer analysis results, our method selected the smallest number of differentially expressed genes and get the highest classification accuracy.

Forecasting Short-Term KOSPI Time Series Based on NEWFM

Fuzzy neural networks have been successfully applied to generate predictive rules for stock forecasting. This paper presents a methodology to forecast the daily Korea composite stock price index (KOSPI) by extracting fuzzy rules based on the neural network with weighted fuzzy membership functions (NEWFM) and the minimized number of input features using the distributed non-overlap area measurement method. NEWFM supports the KOSPI time series analysis based on the defuzzyfication of weighted average method which is the fuzzy model suggested by Takagi and Sugeno. NEWFM classifies upper and lower cases of next daypsilas KOSPI using the recent 32 days of CPPn,m (Current Price Position of day n : a percentage of the difference between the price of day n and the moving average of the past m days from day n-1) of KOSPI. In this paper, the Haar wavelet function is used as a mother wavelet. The most important four input features among CPPn,m and 38 numbers of wavelet transformed coefficients produced by the recent 32 days of CPPn,m are selected by the non-overlap area distribution measurement method. The total number of samples is 2928 trading days, from January 1989 to December 1998. About 80% of the data is used for training and 20% for testing. The result of classification rate is 59.0361%.

A structure-based approach for multimedia information filtering

While multimedia documents are sequentially presented to users, an information filtering (IF) system is useful to achieve a good retrieval performance in terms of both quality and efficiency. Conventional approaches for designing an IF system are based on the users evaluation on information relevance degree (IRD), but ignore other attributes in system design such as relative importance of the data in a collection of multimedia documents. In this paper, we aim at developing a framework of designing structure-based multimedia IF systems, which incorporates the characteristics of the importance and relevance of multimedia documents. A method of calculating the values of relative importance degree of multimedia documents is proposed. Furthermore, these values are combined into the IRD of multimedia documents to improve the representation of user profiles. An illustrative example is given to demonstrate the proposed techniques.

Detection of Ventricular Fibrillation Based on Time Domain Analysis

This study proposes feature extraction using Hilbert transforms and phase space reconstruction to detect ventricular fibrillation (VF) and normal sinus rhythm (NSR) from ECG episodes. We implemented three pre-processing steps to extract features from ECG episodes. In the first step, we use Hilbert transforms to extract peaks. In the second step, we use statistical methods and extract 4 features from the peaks. In the final step, we extract 4 features using statistical methods based on the Euclidean distance between the origin (0, 0) and the peaks after the peaks are plotted in a two dimensional phase space diagram. We applied the 8 features as inputs to a neural network with weighted fuzzy membership functions (NEWFM), and recorded sensitivity, specificity, and accuracy performances of 76.37%, 89.18%, and 86.63%, respectively.