Search

Heartbeat classification plays an important role in identifying cardiac arrhythmias. Although automated heartbeat classification approaches have been broadly reported, they still suffer from some disadvantages, such as the inescapable domain knowledge and human ingenuity for hand-crafted heartbeat feature engineering, and the slow learning speed for iterative deep learning heartbeat classification approach. To address the above issues, for the first time, this paper proposes a novel multilayer extreme learning machine (ML-ELM) heartbeat classification approach for feature representation and classification of the heartbeat signals. Different from the iterative optimization scheme in traditional neural networks, ELM autoencoder (ELM-AE) is employed in ML-ELM for random mapping of the input weights and non-iterative learning is implemented for output weights. To realize unsupervised self-encoding heartbeat feature mapping, the output of ELM-AE is consistent with the incoming heartbeat signal, and the optimization objective is to minimize the reconstruction error. Heartbeat feature extraction is efficiently performed by layer-by-layer ELM-AE stacking. In the stage of heartbeat classification, to solve the unstable performance caused by random parameter initialization in ELM hidden nodes, bagging-based ensemble learning is employed to combine several ELM classifiers and thus develops a well-performed heartbeat classification approach. The proposed approach is applied to two-lead ECG signals, which is obtained from the MIT-BIH arrhythmia public dataset. The experimental results show that the ELM-AE based feature extraction can effectively characterize the characteristics of the heartbeat signal with high efficiency compared with other state-of-the-art approaches. Applying the ensemble decision fusion to two leads, the final classification accuracy reaches 99.41%. [ABSTRACT FROM AUTHOR]

In this world, travelling is the daily routine to reach the desired places such as schools, hospitals, offices and so on. Everyone makes use of their vehicles like two-wheelers or four-wheelers to reach the destination. Vehicle condition plays a major role which includes many parameters such as tyre condition, engine overheating, low fuel mileage. This paper concentrates on the examination of tyre track design for deformities. The mileage of a vehicle relies upon different components. However, it depends on different parameters like street condition, tyre weight and numerous others. For smooth driving and for safe traffic reasons, the auspicious maintenance of tyres plays a vital task. Consequently, the likelihood of prediction of tyre imperfections is assumed to play a vital role to fleet proprietors, government and tyre makers. Researchers make use of recent technologies such as machine learning and deep learning concepts to provide a solution to various real-life problems. The creation of a weighted quality-based convolutional neural network with the purpose of discovery of imperfections on tyres by the examination of pictures of tyre track designs is being proposed. The proposed system with the accessible data-set made public by different tyre producers is included for evaluation. The consequences of our tests demonstrate that it is conceivable to predict the deformities and durability of a tyre from its track designs through a weighted quality-based convolutional neural network. The proposed methodology is used to identify the group of feature values from the tyre image. These kinds of feature values are utilized to implement the dissimilarity of the feature related pixels. The number of epochs is varied as a callback mechanism has been used to validate the proposed model. Binary accuracy and mean absolute error are the best metrics to evaluate the performance which are discussed in the results section. Also, precision, recall and F-measure are used as validation parameters for the proposed model. [ABSTRACT FROM AUTHOR]

Nowadays, big data plays a substantial part in information knowledge analysis, manipulation, and forecasting. Analyzing and extracting knowledge from such big datasets are a very challenging task due to the imbalance of data distribution, which could lead to a biased classification results and wrong decisions. The standard classifiers are not capable of handling such datasets. Hence, a new technique for dealing with such datasets is required. This paper proposes a novel classification framework for big data that consists of three developed phases. The first phase is the feature selection phase, which uses the Whale optimization algorithm (WOA) for finding the best set of features. The second phase is the preprocessing phase, which uses the SMOTE algorithm and the LSH-SMOTE algorithm for solving the class imbalance problem. Lastly, the third phase is WOA + BRNN algorithm, which is using the Whale optimization algorithm for training a deep learning approach called bidirectional recurrent neural network for the first time. Our proposed algorithm WOA-BRNN has been tested against nine highly imbalanced datasets one of them is big dataset in terms of area under curve (AUC) against four of the most common use machine learning algorithms (Naïve Bayes, AdaBoostM1, decision table, random tree), in addition to GWO-MLP (training multilayer perceptron using Gray Wolf Optimizer), then we test our algorithm over four well-known datasets against GWO-MLP and particle swarm optimization (PSO-MLP), genetic algorithm (GA-MLP), ant colony optimization (ACO-MLP), evolution strategy (ES-MLP), and population-based incremental learning (PBIL-MLP) in terms of classification accuracy. Experimental results proved that our proposed algorithm WOA + BRNN has achieved promising accuracy and high local optima avoidance, and outperformed four of the most common use machine learning algorithms, and GWO-MLP in terms of AUC. [ABSTRACT FROM AUTHOR]