MISSING VALUE ESTIMATION METHODS FOR CLASSIFICATION OF ARRHYTHMIA USING DEEP LEARNING
Dipalika Das1*,Maya Nayak2, Subhendu Kumar Pani3
1Dipalika Das, Research Scholar, Department of Computer Science and Engineering, Biju Patnaik University of Technology, Rourkela, Odisha, India; firstname.lastname@example.org,
2Dr. Maya Nayak, Dean School of Computer Studies, Ajay Binay Institute of Technology(ABIT), Cuttack, Biju Patnaik University of Technology(BPUT) Rourkela, Odisha, India; email@example.com,
3Dr. Subhendu Kumar Pani, Professor, Krupajal Engineering College(KEC), Bhubaneswar, Biju Patnaik University of Technology(BPUT) Rourkela, Odisha, India; firstname.lastname@example.org
Abstract: – Biomedical signals like ECG signals are significant to the classification of heart diseases using deep learning techniques. In reality, the ECG datasets mainly consist of matrix data with missing value because of errors or faults. As many classical classification methods, need a full data matrix for input. Therefore, the apt way to impute the missing data is to alleviate the effectiveness of classification of datasets with few missing values. In this paper, the approach of random forest is used for imbalance dataset and compared with other methods e.g. zero method, mean method and PCA based method. The proposed classification algorithm used is Deep Neural Network. The simulation inference is based on the UCI database reflects that random forest method can manage better accuracy while handling missing values in cardiac arrhythmia dataset. Adaptive Neuro-fuzzy inference system classification model works efficiently with proposed method of imputation with efficiency.
Keywords: – Missing Value Estimation, Arrhythmia Classification, Random Forest, Adaptive Neuro fuzzy inference system (ANFIS)