Dong-Kun Shin

Extracting Fuzzy Rules for Detecting Ventricular Arrhythmias Based on NEWFM

In the heart disease, the important problem of ECG arrhythmia is to discriminate ventricular arrhythmias from normal cardiac rhythm. This paper presents novel method based on the neural network with weighted fuzzy membership functions (NEWFM) for the discrimination of ventricular tachycardia (VT) and ventricular fibrillation (VF) from normal sinus rhythm (NSR). This paper uses two pre-processes, the Haar wavelet function and extraction feature method are carried out in order. By using these methods, six features can be generated, which are the input data of NEWFM. NEWFM classifies NSR and VT/VF beats by the trained bounded sum of weighted fuzzy membership functions (BSWFMs) using six input features from the Creighton University Ventricular Tachyarrhythmia Data Base (CUDB). The results are better than Amanns phase space reconstruction (PSR) algorithm, accuracy and specificity rates of 90.4% and 93.3%, respectively.

Wavelet-Based Minimized Feature Selection for Motor Imagery Classification

This paper presents a methodology for classifying left and right motor imagery using a neural network with weighted fuzzy membership functions (NEWFM) and wavelet-based feature extraction. Wavelet coefficients are extracted from electroencephalogram(EEG) signal by wavelet transforms in the first step. In the second step, sixty numbers of initial features are extracted from wavelet coefficients by the frequency distribution and the amount of variability in frequency distribution. The distributed non-overlap area measurement method selects the minimized number of features by removing the worst input features one by one, and then minimized six numbers of features are selected with the highest performance result. The proposed methodology shows that accuracy rate is 86.43% with six numbers of features.

Extracting Fuzzy Rules to Classify Motor Imagery Based on a Neural Network with Weighted Fuzzy Membership Functions

This paper presents a methodology to classify motor imagery by extracting fuzzy rules based on the neural network with weighted fuzzy membership functions (NEWFM) and twenty-four numbers of input features that are extracted by wavelet-based features. This paper consists of three steps to classify motor imagery. In the first step, wavelet transform is performed to filter noises from signals. In the second step, twenty-four numbers of input features are extracted by wavelet-based features from filtered signals by wavelet transform. In the final step, NEWFM classifies motor imagery using twenty-four numbers of input features that are extracted in the second step. In this paper, twenty-four numbers of input features are selected for generating the fuzzy rules to classify motor imagery. NEWFM is tested on the Graz BCI datasets that were used in the BCI Competitions of 2003. The accuracy of NEWFM is 83.51%.

Automated Knowledge Acquisition from Discrete Data Based on NEWFM

A useful technique for automated knowledge acquisition from a database is to select the minimum number of input features with the highest performance result. This paper presents automated knowledge acquisition to using a feature selection based on a neural network with weighted fuzzy membership functions (NEWFM). NEWFM supports the power and usefulness of fuzzy classification rules based on a non-overlap area measurement method. The non-overlap area measurement method selects the minimum number of input features with the highest performance result from initial input features by removing the worst input features one by one. The highest performance results in a non-overlap area distribution measurement method from Credit approval and Australian credit approval at the UCI repository are 87.75% and 87.10%, respectively.

Comparing the Feature Selection Using the Distributed Non-overlap Area Measurement Method with Principal Component Analysis

This paper compares the forecasting performance of the feature extraction using the principal component analysis (PCA) that is one of the oldest and best known techniques in multivariate analysis with the feature selection using the non-overlap area distribution measurement method based on the neural network with weighted fuzzy membership functions (NEWFM). This paper proposes 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) as a new technical indicator. In this paper, two and one input features with the best average forecasting performance are selected from the number of approximations and detail coefficients made by Haar wavelet function from CPPn,5 to CPPn-31,5 using the non-overlap area distribution measurement method and PCA, respectively. The performance results of the non-overlap area distribution measurement method and PCA are 60.93% and 56.63%, respectively. The non-overlap area distribution measurement method outperforms PCA by 4.3% for the holdout sets.

Forecasting KOSPI Using a Neural Network with Weighted Fuzzy Membership Functions and Technical Indicators

This paper presents a methodology to forecast the direction of change in the daily Korea composite stock price index (KOSPI) by extracting fuzzy rules based on the neural network with weighted fuzzy membership functions (NEWFM) and thirteen numbers of input features that are derived by overbought conditions and oversold conditions of three numbers of technical indicators. This paper consists of three steps for forecasting the direction of change in the daily KOSPI. In the first step, three numbers of technical indicators are selected to preprocess the daily KOSPI. In the second step, thirteen numbers of input features are derived by overbought conditions and oversold conditions of three numbers of technical indicators. In the final step, NEWFM classifies the next days direction of change in the daily KOSPI using thirteen numbers of input features that are produced in the second step. The total number of samples is 2928 trading days, from January 1989 to December 1998. About 80% of the whole trading days is used for training and 20% for testing. The performance result of NEWFM for the direction of change in the daily KOSPI is 58.86%.

Detecting Ventricular Tachycardia/Fibrillation Using Neural Network with Weighted Fuzzy Membership Functions and Wavelet Transforms

This paper presents an approach to classify normal and ventricular tachycardia/fibrillation(VT/VF) from the Creighton University Ventricular Tachyarrhythmia Database(CUDB) using the neural network with weighted fuzzy membership functions(NEWFM) and wavelet transforms. In the first step, wavelet transforms are used to obtain the detail coefficients at levels 3 and 4. In the second step, all of detail coefficients d3 and d4 are classified into four intervals, respectively, and then the standard deviations of the specific intervals are used as eight numbers of input features of NEWFM. NEWFM classifies normal and VT/VF beats using eight numbers of input features, and then the accuracy rate is 90.1%.