15.1 Introduction

The necessity for effective monitoring of sub health issues is steadily increasing. Cardiac disease or heart disease, is now the largest cause of death. Heart disease identification is important because it affects not only adults but also youngsters all around the world. It can happen to anyone who has an improper diet, has high cholesterol, smokes, has an alcohol or drug addiction, or is diabetic. An electrocardiograph (ECG) is a quick and straightforward tool to identify, diagnose, and treat cardiac arrhythmia. It’s also not difficult for expert cardiologists to distinguish between dozens of various types of heartbeats. Researchers, on the other hand, have yet to be able to successfully apply state-of-the-art supervised machine learning approaches to attain the same degree of diagnosis.

Cardiac Arrhythmia is a medical term that describes cardiac rhythms that do not follow a regular pattern. A total of 12 different forms of aberrant arrhythmia rhythms exist. The process of finding and classifying arrhythmias can be extremely difficult for a human person since it is often essential to assess each heartbeat of ECG readings obtained by a Holter monitor over hours or even days. Furthermore, due to weariness, there is a risk of human mistake during the processing of ECG records. So, Computational approaches for automatic classification would be a better option.Cardiac Arrhythmia is a medical term that describes cardiac rhythms that do not follow a regular pattern. A total of 12 different forms of aberrant arrhythmia rhythms exist. The process of finding and classifying arrhythmias can be extremely difficult for a human person since it is often essential to assess each heartbeat of ECG readings obtained by a Holter monitor over hours or even days. Furthermore, due to weariness, there is a risk of human mistake during the processing of ECG records. So, Computational approaches for automatic classification would be a better option.

The objective of this article may be summarized:

• To study and examine the arrhythmia classification techniques as practically implementable.
• To overview the existing research studies based on arrhythmia classification benefits and further direction.
• Identify the latest research trends and publications interests based on arrhythmia classification.
• Detection of heart disease in an early stage by using various Deep Learning algorithms on MIT-BIH Arrhythmia dataset from Physionet to train the model.
• To build three different prediction algorithms on collected dataset.
• Also, the validation part of prediction labels will be processed in the testing phase and compared to existing models. After training among our dataset, we would predict normal and abnormal signals.
• To compare the performance of different prediction algorithms.

We are detecting heart arrhythmias using deep learning algorithms so deep learning is stated below.

15.2 Motivation

It is challenging to visually detect heart disease from the electrocardio-graphic (ECG) signals. Implementing an automated ECG signal detection system can aid in the identification of arrhythmia and increase diagnostic accuracy. The main motivation of doing this project is to present a heart disease prediction model for the prediction of occurrence of heart disease. Further, this project work is aimed towards identifying the best classification algorithm for identifying the possibility of heart disease in a patient.It is challenging to visually detect heart disease from the electrocardio-graphic (ECG) signals. Implementing an automated ECG signal detection system can aid in the identification of arrhythmia and increase diagnostic accuracy. The main motivation of doing this project is to present a heart disease prediction model for the prediction of occurrence of heart disease. Further, this project work is aimed towards identifying the best classification algorithm for identifying the possibility of heart disease in a patient.

15.3 Literature Review

Several research articles covering the application of Machine Learning and Data Mining are evaluated in this study in order to acquire an overall understanding of how to deal with the dataset, which algorithms should be used, and how the accuracy can be increased to create an efficient system. The reviews of some of the studies are included below, along with the approach and approach employed.Several research articles covering the application of Machine Learning and Data Mining are evaluated in this study in order to acquire an overall understanding of how to deal with the dataset, which algorithms should be used, and how the accuracy can be increased to create an efficient system. The reviews of some of the studies are included below, along with the approach and approach employed.

Gawande and Barhatte [1] utilized a system, which consist CNN configured in 7 layers consist of input signal, convolution matrix, pooling, stack of line conversion, sigmoid function, another activation function and soft-max function at the output stage. The researchers achieved an accuracy of 99.46%.

Sarvan and Özkurt [2] proposed a model, CNN algorithm has 3 basic layers: a convolutional linear layer, a pooling layer, and fully connected layer. The overall accuracy of the model achieved is 93.72%.

ŞEN and ÖZKURT [3] proposed a CNN model consist of convolutional layer, pooling layer and fully-connected layer. The researchers achieved accuracy for Time-series signal (97.10%) and spectrogram images (99.67%).

Rajkumar et al., [4] proposed approach uses Machine Learning to classify ECG signals in Time Series Analyses. The data is processed through different layer consist of DENSE layer, ELU activation function and Softmax layer. Researchers achieved an accuracy of 93.60%.

Qayyum et al., [7] proposed a 2D CNN model, two different CNN is used one-dimensional and the other is two-dimensional. Even the temporal vector cardiogram (TVCG) approach optimized by the particle swarm optimization (PSG) algorithm and the Support Vector Machine (SVM) with decreased features beat both of these convolutional models. Researchers achieved an accuracy of 2D CNN 94.37% and 1D CNN 92.67%.

Singh et al., [8] proposed CNN model architecture uses a six-layer CNN model with input ECG signal, convolution, pooling, FC, and output layers to generate feature maps. For learning, a stochastic gradient algorithm (SGD) is applied to the CNN architecture. This is a cost function optimization method based on calculative repetition. The overall accuracy of the model is 87.50%

Rana and Kim [9] proposed LSTM networks which are an advancement over typical recurrent neural networks in that they avoid the vanishing gradient problem. Backpropagation through time (BPPT) and its derivatives can be used to train LSTM networks. Tensorflow library’s single layer LSTM model was employed in the methodology. The authors achieved an accuracy of 95%.

Sangeetha et al., [10] proposed CNN which is used as Exponential Linear Units with Xavier initialization for all layers. The suggested CAPE displays the detection of cardiac arrhythmias and the categorization of individuals with similar symptoms in three steps: ECG includes ECG 1D signal to 2D picture conversion, data augmentation and heartbeat classification, and patient grouping based on predictive similarity learning. The overall accuracy of the model is 93.57%.

Takalo-Mattila et al., [12] proposed a system that included a preprocessing unit, ECG windowing, a trained classification model, and compression. Maximum pooling- and dropout-layers are utilized to prevent overfitting. Using extracted features from CNN layers, the MLP classifier is trained to categories ECG beats into five different beat classes, demonstrating that CNN layers may be used as a feature extractor. The authors achieved sensitivity of 92%.

Xu and Liu [13] proposed a method which is composed of two main steps: pre-processing and classification. The raw signal was divided into six levels using the Daubechies wavelet, with the wavelet coefficients from the third to sixth levels being kept and used for reconstruction. Four convolutional layers, two subsampling layers, two fully connected levels, and one Softmax layer make up the CNN classifier’s nine layers. The overall accuracy of the proposed method is 99.43%

Wasimuddin et al., [15] proposed method entails the dataset is collected, the R-R interval of the ECG signal is extracted and represented as a 2-D picture for pre-processing, a CNN classifier is created and trained. The CNN model has ten layers, which include two convolutional layers, the Relu activation function, and finally a pooling layer. The loss function used in this model is cross entropy. The overall accuracy of the proposed method is 97.47%.

Gupta et al., [20] proposed machine intelligence framework for heart disease diagnostics is to maximize the system’s capability in predicting heart disease in order to improve patient survival rates by detecting disease accurately, precisely, and early. constructs Data Imputation and Partitioning, Feature Extraction using FAMD, and Data Imputation and Partitioning are only a few of the phases in MIFH. FAMD also aids in the visualization of object graphical representations, correlation between numeric and categorical data, and feature association. The authors achieved accuracy of 93.44%

Huang et al., [21] have stated that there are two important parameters in the proposed architecture of the 2D-CNN model: the learning rate and the batch size. The phase of model parameter optimization is required to attain the best classification performance of ECG heart rhythm abnormalities. The processing of features in feature-extraction-pattern techniques is substantially more complicated. The overall accuracy of the proposed architecture is 99.0%.

Sarmah [22] proposed methodology consists of 3 steps, Registration, Login and Verification. Huffman coding is a lossless compression strategy based on the frequency of the symbol’s appearance in the file. The patient’s data is encrypted with the Patient Id, Doctor Id, and Hospital Id-Advanced Encryption Standard (PDH-AES) Algorithm. The PDH-AES technology, which is used in secure data transfer, produces the greatest results, with the highest level of security (95.87%) and the quickest encryption and decryption times. The model achieved accuracy of 96.80%.

Kiranyaz et al., [24] proposed approach are in training (offline) and real-time classification and monitoring phases. The raw ECG data from each individual patient in the database is feature extracted and classified using adaptive 1D CNNs. The neurons of the hidden CNN layers are expanded to be capable of both convolution and down-sampling in order to simplify the CNN analogies and have the freedom of any input layer dimension independent of the CNN parameters. The overall accuracy of the proposed approach is VEB – 99.00% and SVEB – 97.60%.

Zhai and Tin (2018) [25] proposed a CNN-based framework for heartbeat classification using dual-beat ECG coupling matrix consist of three convolutional layers, two sub-sampling layers (1 maximum sub-sampling layer and 1 average sub-sampling layer), one fully connected layer with dropout, and a softmax loss layer. The activation function is Rectified linear units (ReLU). To improve classification performance, an automatic selection approach for selecting the most useful training beats was presented. The authors achieved accuracy of VEB – 98.60% and SVEB – 97.50%.

Dang et al., [26] proposed network structure model consists of four convolution layers, two BLSTM layers, two full connection layers, and additional computational activities (pooling layers, ReLU activate, batch normalization, dropout, and so on). Input layer, convolution layer, polling layer, activation function layer, completely connected layer or completely convolution layer, and SoftMax layer are all part of the CNN network model for feature extraction and learning. The overall accuracy of the proposed network structure model is 99.94% (train), 98.63% (validation), and 96.59% (test).

JSonawane and Patil [27] proposed technique for the classification and retrieval of image by using self-organizing map (SOM). In this methodology, the image texture is classified in two phases: the first phase extracts color features, and the second phase uses a self-organizing map to classify the image texture based on the color features. The photos in each class from the first phase are categorized again in the second phase using a self-organizing map based on texture features collected with the GLCM matrix. Multilayer perceptron and other neural networks are trained using the back-propagation algorithm. The model achieved accuracy of 93.39%.

Chauhan et al., [29] proposed a system for heart disease prediction. It analyses a data mining system for properly predicting cardiac disease, which will assist both analysts and doctors. On the patient’s dataset, frequent pattern growth association mining is used to develop strong association rules. Assist doctors in analyzing their data and accurately predicting cardiac problems. The overall accuracy of the proposed a system for heart disease prediction is 61.07% (training) and 53.33% (testing).

Gavhane et al., ([31]) proposed Mechanism used the neural network algorithm multi-layer perceptron (MLP) to train and test the dataset. The multi-layer perceptron algorithm is a supervised neural network approach with one input layer, a second output layer, and one or more hidden layers between these two levels. To obtain the output, the activation function is applied to the weighted input. The input and output layers are connected by the hidden layer, which processes the data internally. The authors achieved precision of 0.91.

Yuwono et al., [32] proposed MKNN (Modified K-Nearest Neighbor) which is a method for recognizing data based on specified ECG component data values such as PR interval, PR segment duration, QRS interval, ST segment duration, QT interval duration, and ST interval duration. There are multiple steps in the MKNN algorithm, including identifying K, computing validity, computing Euclidean distance, calculating the weights and determination of data classes evaluated based on selected k. The overall accuracy of the proposed methodology is 71.20%.

Ambekar and Phalnikar [33] proposed a system for classification which are nave Bayes and KNN. Researchers can estimate if a patient is at high or low risk using the CNN-UDRP algorithm. The Nave Bayes classifier provides input to CNN-UDRP (highest accuracy value as compared to KNN classifier). The CNN algorithm is used to extract features. The SoftMax classifier is used to classify the risk of heart disease. Researchers achieved accuracy of 82%.

Suvarna et al., [35] proposed system uses CPSO (Constricted Particle Swarm Optimization) and PSO (Particle Swarm Optimization) for prediction of heart disease. The proposed prediction algorithm in action. A particle evaluates its own experience, which is specified by the Euclidean distance between two particles positions, or as a sociometric neighbor-hood, while selecting where to move next. The overall accuracy of the proposed system is 53.1%.

Singh et al., [37] proposed technique generates Classification Association Rules (CARs) and determine which approach provides the highest percentage of accurately projected values for early heart disease diagnosis. The proposed method was compared to existing state-of-the-art procedures in a comparative analysis. Selection, Pre-processing and Transformation, Selection of Associative rules, Performance Evaluation, and Predict Diseases the IHDPS (Intelligent Heart Disease Prediction System) is designed to detect heart disorders based on improved performance and correctly identified cases of the applied algorithm. The model achieved accuracy of 99.19%.

Ramprakash et al., [39] used the neural network acquired a training dataset. Pattern recognition is accomplished using neural networks, which are a set of algorithms. Activation functions make up the neural network’s layers. The number of guidelines used to determine the network’s behav-ior has an impact on its performance. Low capacity occurs from a model with fewer parameters, resulting in underfitting. Overfitting occurs when a model contains more parameters than is required. The overall accuracy of the proposed methodology is 94%.

Nikhar and Karandikar [40] employed the Decision Tree Classification method and the Nave Bayes Classifier algorithm. Both category and numerical data are handled by the Decision Tree Classification method. It provides a categorical solution, such as Yes/No, based on specified conditions. The Decision tree Classification technique is commonly used to handle medical datasets. The most efficient and economical classification technique is the Nave Bayes Classifier Algorithm, which can handle vast, difficult, non-linear, dependent data. The authors achieved accuracy of 91% (DT model) and 87% (NB model).

Jangir et al., [41], Choubey et al., [42, 43, 50, 52, 54, 67, 68] have used similar data science and machine learning algorithms for the identifications and predictions of medical diabetes. The idea conceived through the review [49, 53, 57, 59–61, 64, 66] of many published articles, text and references like comparative analysis of classification methods, performance evaluation of classification methods [44], rule based diagnosis system [45] and classification techniques diagnosis [46–48, 58] for diabetes, classification techniques diagnosis [51] for leukemia, classification techniques diagnosis [55, 56] for heart disease, classification techniques diagnosis [62] for dengue, image detection [63] using computer vision are found to be of great help in accomplishment of the present work.

Table 15.1 consists of the summary of existing work for heart disease.

Table 15.2 consists of the used techniques with their advantages and issues.

Table 15.3 consists of the summary of future works over exiting works.

In this chapter various studies were reviewed on the basis of the technique used, tool used, algorithm used and thereby ending with the conclusion. These studies helped the authors to get prepared for the upcoming situation.

15.4 Proposed Approach

We started by making a list of all the patients in the data path. We used a pypi package wfdb for loading the ECG and annotations and made a list of the non-beat and abnormal beats. Then we made a dataset that is cen-tered on beats with +- 3 seconds before and after. We then split our data by patients into a train and validation set because technically the same patient can show up in both the training and validation sets. This means that we may accidentally leak information across the datasets. Therefore, we split on patients instead of samples.

The following software and hardware are required for the designed algorithms which is mentioned below: (a) Software: Ubuntu 16 or above/ Windows 7 or above, Python, Packages (Numpy, Sklearn, Matplotlib, Pandas, Numpy, Wfdb), Jupyter Notebook, Sublime Text, Google Chrome (b) Hardware: 4 GB RAM or above, 1 TB ROM or above, Processor Speed 1.4 GHz or above.

The dataflow diagram for the proposed approach is shown below in Figure 15.1.

To prevent the model from overfitting and tracing the training process, 70% of the training data is actually used to train the model, while 30% of the training data is used to validate the performance of the network at the end of each epoch. The model is trained for 5 epochs using the Dense NN, CNN and LSTM algorithm with a batch size of 32. The learning rate is set to 0.001 and is also applied to Adam optimizer so as to accelerate the learningto 0.001 and is also applied to Adam optimizer so as to accelerate the learning process. The total training and validation process was processed under the GPU acceleration environment.

The work descriptions deal into two stages: (a) Dataset (b) Algorithms

15.4.2 Algorithms Description

In this study, we used Jupyter Notebook to determine the prediction of arrhythmia using Python. It has inbuild libraries to do the statistical computation in few seconds.

Figure 15.2 presents the proposed architecture.

The detailed description of the algorithms used in the model are given below.

15.4.2.1 Dense Neural Network

A deep neural network (DNN) is an artificial neural network (ANN) with multiple layers between the input and output layers There are different types of neural networks but they always consist of the same components: neurons, synapses, weights, biases, and functions. These components functioning similar to the human brains and can be trained like any other ML algorithm.

DNNs are typically feedforward networks in which data flows from the input layer to the output layer without looping back. At first, the DNN creates a map of virtual neurons and assigns random numerical values, or “weights”, to connections between them. The weights and inputs are multiplied and return an output between 0 and 1. If the network did not accurately recognize a particular pattern, an algorithm would adjust the weights. That way the algorithm can make certain parameters more influential, until it determines the correct mathematical manipulation to fully process the data.

Dense NN-pseudo-code involves training of the required dataset and after that, evaluation of the model is performed on the test dataset. Finally, accuracy, recall, precision, specificity, and prevalence will be the output of the model.

The Algorithm 15.1 of DNN is as noted below.

Figure 15.3 depicts the DNN architecture.

15.4.2.2 Convolutional Neural Network

A Convolution Neural Network (CNN) is a special type of deep learning algorithm which uses a set of filters and the convolution operator to reduce the number of parameters. This algorithm sparked the state-of-the-art techniques for image classification. Essentially, the way this works for 1D CNN is to take a filter (kernel) of size kernel_size starting with the first-time stamp. The convolution operator takes the filter and multiplies each element against the first kernel_size time steps. These products are then summed for the first cell in the next layer of the neural network.

The filter then moves over by stride time steps and repeats. The default stride in keras is 1, which we will use. In image classification, most people use padding which allows you to pick up some features on the edges of the image by adding ‘extra’ cells, we will use the default padding which is 0. The output of the convolution is then multiplied by a set of weights W and added to a bias b and then passed through a non-linear activation function as in a dense neural network. We can then repeat this with addition CNN layers if desired. Here we will use Dropout which is a technique for reducing overfitting by randomly removing some nodes.

CNN-pseudo-code involves training of the required dataset and after that, evaluation of the model is performed on the test dataset. Finally, accuracy, recall, precision, specificity, and prevalence will be the output of the model.

The Algorithm 15.2 of CNN is as noted below.

Figure 15.4 presents the CNN architecture.

15.4.2.3 Long Short-Term Memory

Long short-term memory (LSTM) is an artificial recurrent neural network (RNN) architecture used in the field of deep learning. Unlike standard feed-forward neural networks, LSTM has feedback connections. It can not only process single data points (such as images), but also entire sequences of data (such as speech or video). For example, LSTM is applicable to tasks such as unsegmented, connected handwriting recognition, speech recognition and anomaly detection in network traffic or IDSs (intrusion detection systems).

Since this data signal is time-series, it is natural to test a recurrent neural network (RNN). Here we will test a bidirectional long short-term memory (LSTM). Unlike in dense NN and CNN, RNN have loops in the network to keep a memory of what has happened in the past. This allows the network to pass information from early time steps to later time steps that usually would be lost in other types of networks. Essentially there is an extra term for this memory state in the calculation before passing through a non-linear activation function. Here we use the bidirectional information so information can be passed in both direction (left to right and right to left). This will help us pick up information about the normal heart beats to the left and right of the center heartbeat.

Figure 15.5 shows the LSTM network.

15.5 Experimental Results of Proposed Approach

We have used DNN, CNN and LSTM on MIT-BIH Arrhythmia Database. The performance has been evaluated in tabular forms.

Table 15.5 presents the training performance of DNN, CNN and LSTM.

The above table represents the training parameter values such as AUC, Accuracy, Recall, Precision, Specificity and Prevalence of DNN, CNN, and LSTM.

Table 15.6 presents the testing performance of DNN, CNN and LSTM. The above table represents the testing parameter values such as AUC, Accuracy, Recall, Precision, Specificity and Prevalence of DNN, CNN, and LSTM.

Table 15.7 presents the comparison of proposed algorithms with existing in context of accuracy.

In above table, it may be observed that proposed algorithm i.e., DNN achieved the best performance than existing.

Figure 15.6 shows the ECG Signal vs. Time Index of Abnormal Beats.

Figure 15.7 shows AUC vs. Number Training Points.

Figure 15.8 shows the ROC curve of DNN, CNN and LSTM.

Figure 15.9 shows the ROC curve of LSTM.

Figure 15.10 shows validation of CNN, DNN, and LSTM.

15.6 Conclusion and Future Scope

More data set can be taken in consideration with optimized hyperparam-eters in order to achieve even higher values for CNN, Dense CNN and LSTM network.

Determining any heart disease on some raw data is really difficult for even a doctor which is why many healthcare sectors are opting for machine learning techniques to determine it. In our project we took the MIT-BIH Arrhythmia dataset from https://physionet.org/content/mitdb/1.0.0/ and applied pre-processing to drop the data with missing values and applied some deep learning models like Dense neural network, CNN, LSTM of which Dense NN gives an accuracy of 96.4%, CNN gives an accuracy of 81.2%, and LSTM gives an accuracy of 56.5%. However, the accuracy can be improved by using some data mining techniques for feature extraction from the samples. The ROC curve for Dense NN gives an AUC of 0.866, CNN gives an AUC of 0.907, and LSTM gives an AUC of 0.564.

For further implementation, we can try to take in consideration of huge data set for model training and validation, further more optimize the hyperparameters or number of layers. More data set can be taken in consideration as suggested by learning curve. Instead of taking in consideration of 6 second window centered on the peak of the heartbeat, it can be increased (keeping in mind the degree of complexity of handling the huge information).

References

1. 1. Gawande, N. and Barhatte, A., Heart diseases classification using con-volutional neural network, in: 2017 2nd International Conference on Communication and Electronics Systems (ICCES), 2017, October, IEEE, pp. 17–20.
2. 2. Sarvan, Ç. and Özkurt, N., ECG beat arrhythmia classification by using 1-D CNN in case of class imbalance, in: 2019 Medical Technologies Congress (TIPTEKNO), 2019, October, IEEE, pp. 1–4.
3. 3. ŞEN, S.Y. and Özkurt, N., ECG arrhythmia classification by using convolutional neural network and spectrogram, in: 2019 Innovations in Intelligent Systems and Applications Conference (ASYU), 2019, October, IEEE, pp. 1–6.
4. 4. Rajkumar, A., Ganesan, M., Lavanya, R., Arrhythmia classification on ECG using deep learning, in: 2019 5th International Conference on Advanced Computing & Communication Systems (ICACCS), 2019, March, IEEE, pp. 365–369.
5. 5. Dang, H., Sun, M., Zhang, G., Zhou, X., Chang, Q., Xu, X., A novel deep con-volutional neural network for arrhythmia classification, in: 2019 International Conference on Advanced Mechatronic Systems (ICAMechS), 2019, August, IEEE, pp. 7–11.
6. 6. Hassan, S.U., Zahid, M.S.M., Husain, K., Performance comparison of CNN and LSTM algorithms for arrhythmia classification, in: 2020 International Conference on Computational Intelligence (ICCI), 2020, October, IEEE, pp. 223–228.
7. 7. Qayyum, A.B.A., Islam, T., Haque, M.A., ECG heartbeat classification: A comparative performance analysis between one and two dimensional convolutional neural network, in: 2019 IEEE International Conference on Biomedical Engineering, Computer and Information Technology for Health (BECITHCON), 2019, November, IEEE, pp. 93–96.
8. 8. Singh, S., Sunkaria, R.K., Saini, B.S., Kumar, K., Atrial fibrillation and premature contraction classification using convolutional neural network, in: 2019 International Conference on Intelligent Computing and Control Systems (ICCS), 2019, May, IEEE, pp. 797–800.
9. 9. Rana, A. and Kim, K.K., ECG heartbeat classification using a single layer LSTM model, in: 2019 International SoC Design Conference (ISOCC), 2019, October, IEEE, pp. 267–268.
10. 10. Sangeetha, D., Selvi, S., Ram, M.S.A., A CNN based similarity learning for cardiac arrhythmia prediction, in: 2019 11th International Conference on Advanced Computing (ICoAC), 2019, December, IEEE, pp. 244–248.
11. 11. Prawira, R.H., Wibowo, A., Yusuf, A.Y.P., Best parameters selection of arrhythmia classification using convolutional neural networks, in: 2019 3rd International Conference on Informatics and Computational Sciences (ICICoS), 2019, October, IEEE, pp. 1–6.
12. 12. Takalo-Mattila, J., Kiljander, J., Soininen, J.P., Inter-patient ECG classifi-cation using deep convolutional neural networks, in: 2018 21st Euromicro Conference on Digital System Design (DSD), 2018, August, IEEE, pp. 421–425.
13. 13. Xu, X. and Liu, H., ECG heartbeat classification using convolutional neural networks. IEEE Access, 8, 8614–8619, 2020.
14. 14. Kido, K., Ono, N., Altaf-Ul-Amin, M.D., Kanaya, S., Huang, M., The feasibility of arrhythmias detection from a capacitive ECG measurement using convolutional neural network, in: 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2019, July, IEEE, pp. 3494–3497.
15. 15. Wasimuddin, M., Elleithy, K., Abuzneid, A., Faezipour, M., Abuzaghleh, O., ECG signal analysis using 2-D image classification with convolutional neural network, in: 2019 International Conference on Computational Science and Computational Intelligence (CSCI), 2019, December, IEEE, pp. 949–954.
16. 16. Shaker, A.M., Tantawi, M., Shedeed, H.A., Tolba, M.F., Generalization of convolutional neural networks for ECG classification using generative adversarial networks. IEEE Access, 8, 35592–35605, 2020.
17. 17. Fitriyani, N.L., Syafrudin, M., Alfian, G., Rhee, J., HDPM: An effective heart disease prediction model for a clinical decision support system. IEEE Access, 8, 133034–133050, 2020.
18. 18. Mohan, S., Thirumalai, C., Srivastava, G., Effective heart disease prediction using hybrid machine learning techniques. IEEE Access, 7, 81542–81554, 2019.
19. 19. Li, P., Wang, Y., He, J., Wang, L., Tian, Y., Zhou, T.S., Li, J.S., High-performance personalized heartbeat classification model for long-term ECG signal. IEEE Trans. Biomed. Eng., 64, 1, 78–86, 2016.
20. 20. Gupta, A., Kumar, R., Arora, H.S., Raman, B., MIFH: A machine intelligence framework for heart disease diagnosis. IEEE Access, 8, 14659–14674, 2019.
21. 21. Huang, J., Chen, B., Yao, B., He, W., ECG arrhythmia classification using STFT-based spectrogram and convolutional neural network. IEEE Access, 7, 92871–92880, 2019.
22. 22. Sarmah, S.S., An efficient IoT-based patient monitoring and heart disease prediction system using deep learning modified neural network. IEEE Access, 8, 135784–135797, 2020.
23. 23. Cai, J., Sun, W., Guan, J., You, I., Multi-ECGNet for ECG arrythmia multi- label classification. IEEE Access, 8, 110848–110858, 2020.
24. 24. Kiranyaz, S., Ince, T., Gabbouj, M., Real-time patient-specific ECG classification by 1-D convolutional neural networks. IEEE Trans. Biomed. Eng., 63, 3, 664–675, 2015.
25. 25. Zhai, X. and Tin, C., Automated ECG classification using dual heartbeat coupling based on convolutional neural network. IEEE Access, 6, 27465–27472, 2018.
26. 26. Dang, H., Sun, M., Zhang, G., Qi, X., Zhou, X., Chang, Q., A novel deep arrhythmia-diagnosis network for atrial fibrillation classification using electrocardiogram signals. IEEE Access, 7, 75577–75590, 2019.
27. 27. Sonawane, J.S. and Patil, D.R., Prediction of heart disease using multilayer perceptron neural network, in: International Conference on Information Communication and Embedded Systems (ICICES2014), 2014, February, IEEE, pp. 1–6.
28. 28. Babu, S., Vivek, E.M., Famina, K.P., Fida, K., Aswathi, P., Shanid, M., Hena, M., Heart disease diagnosis using data mining technique, in: 2017 International Conference of Electronics, Communication and Aerospace Technology (ICECA), 2017, April, vol. 1, IEEE, pp. 750–753.
29. 29. Chauhan, A., Jain, A., Sharma, P., Deep, V., Heart disease prediction using evolutionary rule learning, in: 2018 4th International Conference on Computational Intelligence & Communication Technology (CICT), 2018, February, IEEE, pp. 1–4.
30. 30. Khemphila, A. and Boonjing, V., Heart disease classification using neural network and feature selection, in: 2011 21st International Conference on Systems Engineering, 2011, August, IEEE, pp. 406–409.
31. 31. Gavhane, A., Kokkula, G., Pandya, I., Devadkar, K., Prediction of heart disease using machine learning, in: 2018 Second International Conference on Electronics, Communication and Aerospace Technology (ICECA), 2018, March, IEEE, pp. 1275–1278.
32. 32. Yuwono, T., Franz, A., Muhimmah, I., Design of smart electrocardiography (ECG) using Modified K-Nearest Neighbor (MKNN), in: 2018 1st International Conference on Computer Applications & Information Security (ICCAIS), 2018, April, IEEE, pp. 1–5.
33. 33. Ambekar, S. and Phalnikar, R., Disease risk prediction by using convolutional neural network, in: 2018 Fourth International Conference on Computing Communication Control and Automation (ICCUBEA), 2018, August, IEEE, pp. 1–5.
34. 34. Chakarverti, M., Yadav, S., Rajan, R., Classification technique for heart disease prediction in data mining, in: 2019 2nd International Conference on Intelligent Computing, Instrumentation and Control Technologies (ICICICT), 2019, July, vol. 1, IEEE, pp. 1578–1582.
35. 35. Suvarna, C., Sali, A., Salmani, S., Efficient heart disease prediction system using optimization technique, in: 2017 International Conference on Computing Methodologies and Communication (ICCMC), 2017, July, IEEE, pp. 374–379.
36. 36. Dharmasiri, N.D.K.G. and Vasanthapriyan, S., Approach to heart diseases diagnosis and monitoring through machine learning and iOS mobile application, in: 2018 18th International Conference on Advances in ICT for Emerging Regions (ICTer), 2018, September, IEEE, pp. 407–412.
37. 37. Singh, J., Kamra, A., Singh, H., Prediction of heart diseases using associative classification, in: 2016 5th International Conference on Wireless Networks and Embedded Systems (WECON), 2016, October, IEEE, pp. 1–7.
38. 38. Shouman, M., Turner, T., Stocker, R., Using data mining techniques in heart disease diagnosis and treatment, in: 2012 Japan-Egypt Conference on Electronics, Communications and Computers, 2012, March, IEEE, pp. 173–177.
39. 39. Ramprakash, P., Sarumathi, R., Mowriya, R., Nithyavishnupriya, S., Heart disease prediction using deep neural network, in: 2020 International Conference on Inventive Computation Technologies (ICICT), 2020, February, IEEE, pp. 666–670.
40. 40. Nikhar, S. and Karandikar, A.M., Prediction of heart disease using machine learning algorithms. Int. J. Adv. Eng. Manage. Sci., 2, 6, 239484, 2016.
41. 41. Jangir, S.K., Joshi, N., Kumar, M., Choubey, D.K., Singh, S., Verma, M., Functional link convolutional neural network for the classification of diabetes mellitus. Int. J. Numer. Methods Biomed. Eng., e3496, 37, 8, 1–12, 2021.
42. 42. Choubey, D.K., Tripathi, S., Kumar, P., Shukla, V., Dhandhania, V.K., Classification of diabetes by Kernel based SVM with PSO. Recent Adv. Comput. Sci. Commun. (Formerly: Recent Patents Comput. Science), 14, 4, 1242–1255, 2021.
43. 43. Choubey, D.K., Kumar, M., Shukla, V., Tripathi, S., Dhandhania, V.K., Comparative analysis of classification methods with PCA and LDA for diabetes. Curr. Diabetes Rev., 16, 8, 833–850, 2020.
44. 44. Choubey, D.K., Kumar, P., Tripathi, S., Kumar, S., Performance evaluation of classification methods with PCA and PSO for diabetes. Netw. Model. Anal. Health Inform. Bioinform., 9, 1, 1–30, 2020.
45. 45. Choubey, D.K., Paul, S., Dhandhenia, V.K., Rule based diagnosis system for diabetes. Int. J. Med. Sci., 28, 12, 5196–5208, 2017.
46. 46. Choubey, D.K. and Paul, S., GA_RBF NN: A classification system for diabetes. Int. J. Biomed. Eng. Technol., 23, 1, 71–93, 2017.
47. 47. Choubey, D.K. and Paul, S., Classification techniques for diagnosis of diabetes: A review. Int. J. Biomed. Eng. Technol., 21, 1, 15–39, 2016.
48. 48. Choubey, D.K. and Paul, S., GA_SVM: A classification system for diagnosis of diabetes, in: Handbook of Research on Soft Computing and Nature-Inspired Algorithms, pp. 359–397, IGI Global, USA, 2017.
49. 49. Bala, K., Choubey, D.K., Paul, S., Lala, M.G.N., Classification techniques for thunderstorms and lightning prediction: A survey, in: Soft-Computing-Based Nonlinear Control Systems Design, pp. 1–17, IGI Global, USA, 2018.
50. 50. Choubey, D.K., Paul, S., Bala, K., Kumar, M., Singh, U.P., Implementation of a hybrid classification method for diabetes, in: Intelligent Innovations in Multimedia Data Engineering and Management, pp. 201–240, IGI Global, USA, 2019.
51. 51. Rawal, K., Parthvi, A., Choubey, D.K., Shukla, V., Prediction of leukemia by classification and clustering techniques, in: Machine Learning, Big Data, and IoT for Medical Informatics, pp. 275–295, Academic Press, UK, 2021.
52. 52. Choubey, D.K., Paul, S., Kumar, S., Kumar, S., Classification of Pima Indian diabetes dataset using naive Bayes with genetic algorithm as an attribute selection, in: Communication and Computing Systems: Proceedings of the International Conference on Communication and Computing System (ICCCS 2016), 2017, February, pp. 451–455.
53. 53. Bala, K., Choubey, D.K., Paul, S., Soft computing and data mining techniques for thunderstorms and lightning prediction: A survey, in: 2017 International Conference of Electronics, Communication and Aerospace Technology (ICECA), 2017, April, vol. 1, IEEE, pp. 42–46.
54. 54. Choubey, D.K., Paul, S., Dhandhania, V.K., GA_NN: An intelligent classifi-cation system for diabetes, in: Soft Computing for Problem Solving, pp. 11–23, Springer, Singapore, 2019.
55. 55. Kumar, S., Mohapatra, U.M., Singh, D., Choubey, D.K., IoT-based cardiac arrest prediction through heart variability analysis. Advanced Computing and Intelligent Engineering: Proceedings of ICACIE, 2018, vol. 2, p. 353, 2020.
56. 56. Kumar, S., Mohapatra, U.M., Singh, D., Choubey, D.K., EAC: Efficient associative classifier for classification, in: 2019 International Conference on Applied Machine Learning (ICAML), 2019, May, IEEE, pp. 15–20.
57. 57. Pahari, S. and Choubey, D.K., Analysis of liver disorder using classification techniques: A survey, in: 2020 International Conference on Emerging Trends in Information Technology and Engineering (ic-ETITE), 2020, February, IEEE, pp. 1–4.
58. 58. Kumar, S., Bhusan, B., Singh, D., Kumar Choubey, D., Classification of diabetes using deep learning, in: 2020 International Conference on Communication and Signal Processing (ICCSP), 2020, July, IEEE, pp. 0651–0655.
59. 59. Parthvi, A., Rawal, K., Choubey, D.K., A comparative study using machine learning and data mining approach for leukemia, in: 2020 International Conference on Communication and Signal Processing (ICCSP), 2020, July, IEEE, pp. 0672–0677.
60. 60. Sharma, D., Jain, P., Choubey, D.K., A comparative study of computational intelligence for identification of breast cancer, in: International Conference on Machine Learning, Image Processing, Network Security and Data Sciences, 2020, July, Springer, Singapore, pp. 209–216.
61. 61. Srivastava, K. and Choubey, D.K., Soft computing, data mining, and machine learning approaches in detection of heart disease: A review, in: International Conference on Hybrid Intelligent Systems, 1179, pp. 165–175, Springer, Cham, 2021.
62. 62. Choubey, D.K., Mishra, A., Pradhan, S.K., Anand, N., Soft computing techniques for dengue prediction, in: 2021 10th IEEE International Conference on Communication Systems and Network Technologies (CSNT), 2021, June, IEEE, pp. 648–653.
63. 63. Bhatia, U., Kumar, J., Choubey, D.K., Drowsiness image detection using computer vision, in: Soft Computing: Theories and Applications, pp. 667–683, Springer, Singapore, 2022.
64. 64. Pahari, S. and Choubey, D.K., Analysis of liver disorder by machine learning techniques, in: Soft Computing: Theories and Applications, pp. 587–601, Springer, Singapore, 2022.
65. 65. Choubey, D.K., Mishra, A., Pradhan, S.K., Anand, N., Soft computing techniques for dengue prediction, in: 2021 10th IEEE International Conference on Communication Systems and Network Technologies (CSNT), 2021, June, IEEE, pp. 648–653.
66. 66. Choubey, D.K., Paul, S., Bhattacharjee, J., Soft computing approaches for diabetes disease diagnosis: A survey. Int. J. Appl. Eng. Res., 9, 21, 11715–11726, 2014.
67. 67. Choubey, D.K. and Paul, S., GA_J48graft DT: A hybrid intelligent system for diabetes disease diagnosis. Int. J. Bio-Sci. Bio-Technol., 7, 5, 135–150, 2015.
68. 68. Choubey, D.K. and Paul, S., GA_MLP NN: A hybrid intelligent system for diabetes disease diagnosis. Int. J. Intelligent Syst. Appl., 8, 1, 49, 2016.