Application of Deep Learning Techniques in Diagnosis of Covid-19 (Coronavirus): A Systematic Review

The automatic recognition of COVID-19 diseases is critical in the present pandemic since it relieves healthcare staff of the burden of screening for infection with COVID-19. Previous studies have proven that deep learning algorithms can be utilized to aid in the diagnosis of patients with potential COVID-19 infection. However, the accuracy of current COVID-19 recognition models is relatively low. Motivated by this fact, we propose three deep learning architectures, F-EDNC, FC-EDNC, and O-EDNC, to quickly and accurately detect COVID-19 infections from chest computed tomography (CT) images. Sixteen deep learning neural networks have been modified and trained to recognize COVID-19 patients using transfer learning and 2458 CT chest images. The proposed EDNC has then been developed using three of sixteen modified pre-trained models to improve the performance of COVID-19 recognition. The results suggested that the F-EDNC method significantly enhanced the recognition of COVID-19 infections with 97.75% accuracy, followed by FC-EDNC and O-EDNC (97.55% and 96.12%, respectively), which is superior to most of the current COVID-19 recognition models. Furthermore, a localhost web application has been built that enables users to easily upload their chest CT scans and obtain their COVID-19 results automatically. This accurate, fast, and automatic COVID-19 recognition system will relieve the stress of medical professionals for screening COVID-19 infections.

https://www.mdpi.com/2379-139X/8/2/71/pdf?version=1647860194

EDNC: Ensemble Deep Neural Network for COVID-19 Recognition

Stock market forecasting using a multi-task approach integrating long short-term memory and the random forest framework

Cryptocurrencies can be considered as mathematical money. As the most famous cryptocurrency, the Bitcoin price forecasting model is one of the popular mathematical models in financial technology because of its large price fluctuations and complexity. This paper proposes a novel ensemble deep learning model to predict Bitcoin’s next 30 min prices by using price data, technical indicators and sentiment indexes, which integrates two kinds of neural networks, long short-term memory (LSTM) and gate recurrent unit (GRU), with stacking ensemble technique to improve the accuracy of decision. Because of the real-time updates of comments on social media, this paper uses social media texts instead of news websites as the source data of public opinion. It is processed by linguistic statistical method to form the sentiment indexes. Meanwhile, as a financial market forecasting model, the model selects the technical indicators as input as well. Real data from September 2017 to January 2021 is used to train and evaluate the model. The experimental results show that the near-real time prediction has a better performance, with a mean absolute error (MAE) 88.74% better than the daily prediction. The purpose of this work is to explain our solution and show that the ensemble method has better performance and can better help investors in making the right investment decision than other traditional models.

/pdf/a-stacking-ensemble-deep-learning-model-for-bitcoin-price-bdeyi33x.pdf

A Stacking Ensemble Deep Learning Model for Bitcoin Price Prediction Using Twitter Comments on Bitcoin

Text Classification is an important research area in natural language processing (NLP) that has received a considerable amount of scholarly attention in recent years. However, real Chinese online news is characterized by long text, a large amount of information and complex structure, which also reduces the accuracy of Chinese long text classification as a result. To improve the accuracy of long text classification of Chinese news, we propose a BERT-based local feature convolutional network (LFCN) model including four novel modules. First, to address the limitation of Bidirectional Encoder Representations from Transformers (BERT) on the length of the max input sequence, we propose a named Dynamic LEAD-n (DLn) method to extract short texts within the long text based on the traditional LEAD digest algorithm. In Text-Text Encoder (TTE) module, we use BERT pretrained language model to complete the sentence-level feature vector representation of a news text and to capture global features by using the attention mechanism to identify correlated words in text. After that, we propose a CNN-based local feature convolution (LFC) module to capture local features in text, such as key phrases. Finally, the feature vectors generated by the different operations over several different periods are fused and used to predict the category of a news text. Experimental results show that the new method further improves the accuracy of long text classification of Chinese news. 

A Long-Text Classification Method of Chinese News Based on BERT and CNN

Cryptocurrency market has witnessed a boom during the global pandemic and has proven as a strong investment with a wide institutional adoption. A time-series forecasting solution will play a vital role in analyzing the fluctuation of the bitcoin and altcoin markets. Dogecoin is one such altcoin that is a low-price, high-risk investment option garnering considerable interest this year. The variation of the price trend of this altcoin is studied using the multi-head attention mechanism implemented in a transformer, where the attention heads attend to the tokens that are relevant to each current token based on varying short-term and long-term dependencies. In this paper, a multi-head attention-based transformer encoder-decoder model is applied on the hourly data of the Dogecoin price for its prediction over time. The performance of the model has been evaluated using a number of evaluation metrics including MAE and predictive R-squared value. The model trained over the Dogecoin hourly price variation gave an impressive accuracy of 98.46% and R-squared value of 0.8616 comparable with the existing state-of-the-art cryptocurrency price forecasting models.

Multi-Head Self-Attention Transformer for Dogecoin Price Prediction

The stock market is a dynamic and volatile platform which provides an environment for traders to invest and trade in shares. The price of a stock is dependent on numerous static and dynamic features. Predicting the future price of a particular company’s stock can be extremely beneficial for traders. Seq2Seq modelling helps map an input sequence to an output sequence. In this paper, we propose a system to predict the future Open, High, Close, Low (OHCL) value of a stock using a Bi-Directional LSTM based Sequence to Sequence Modelling. Each OHCL price is an independent sequence and multitask learning helps map the interrelations between them. A multitask system is also proposed which uses sub tasks and shared tasks to model the prices. Stock prices of Tata Consumer Products Limited from the National Stock Exchange (NSE) of India is used. To evaluate the efficiency of the proposed systems, they are compared against various machine learning algorithms. The proposed Seq2Seq and multitask systems comfortably outperform the existing algorithms with RMSE values of 3.98 and 7.87 respectively.

Stock Price Prediction using Bi-Directional LSTM based Sequence to Sequence Modeling and Multitask Learning

For an aspiring graduate student, shortlisting the universities to apply to is a difficult problem. Since an application is extremely dynamic, students often tend to wonder if their profile matches the requirement of a certain university. Moreover, the cost of applying to a university is extremely high making it critical that students shortlist universities based on their profile. A university admission prediction system is quite useful for students to determine their chances of acceptance to a specific university. The system could make use of data related to previous applicants to various universities and their admit or reject status. Earlier models of such prediction systems suffer from several drawbacks such as not considering important parameters like GRE (Graduate Record Exam) scores or research experience. Further, the accuracy reported by earlier models is also not sufficiently high. In this paper, a stacked ensemble model that predicts the chances of admit of a student to a particular university has been proposed. The proposed model takes into consideration various factors related to the student including their research experience, industry experience etc. Further, the system proposed has been evaluated against various other machine learning algorithms including other deep learning methods. It is observed that the proposed model easily outperforms all other models and provides a very high accuracy.

A University Admission Prediction System using Stacked Ensemble Learning

The occurrence of imbalanced datasets in medical imaging has proven to be a challenge for the development of models to analyze and evaluate the underlying condition. In this paper, the bias of the chest CT scan dataset is handled by taking discrete splits and employing ResNets to detect COVID-19 in each split. The scraped images were pre-processed using CLAHE histogram for comparison with low contrast images. Multiple ResNets were extended to form an ensemble neural network model using ANNs which handles the class imbalance. The system has an overall accuracy of 87.23% and the performance is assessed for each class. The image features identified are visualized using the GradCAM algorithm and some of the commonly found clinical features in the CT scan images of the patients suffering from this disease are summarized for better understanding the working of the model.

/pdf/covid-19-identification-in-clahe-enhanced-ct-scans-with-34whu0ur0e.pdf

COVID-19 Identification in CLAHE Enhanced CT Scans with Class Imbalance using Ensembled ResNets

The identification of failures and defects in industrial machines has proven to be a challenge to gauge their warranty and performance. Depreciation in industrial machines occurs due to several factors, most commonly- tool wear, strain, heat and power failure. In this paper, the development of machine learning algorithms for the prediction of machine failures is done. A synthesized dataset was used in the predictive maintenance model, that reflects real-time failures encountered in the industries. The class data imbalance hinders the performance of machine learning algorithms and this is handled by evaluating SMOTE-based oversampling techniques. By using SMOTE technique, a 7.83 % increase in the AUC score is observed, thereby improving the performance of the Random Forest classifier in distinguishing the instances of non-failure and machine failures.

Sashank Sridhar

Papers

Multi-Head Self-Attention Transformer for Dogecoin Price Prediction

Stock Price Prediction using Bi-Directional LSTM based Sequence to Sequence Modeling and Multitask Learning

A University Admission Prediction System using Stacked Ensemble Learning

COVID-19 Identification in CLAHE Enhanced CT Scans with Class Imbalance using Ensembled ResNets

Handling Data Imbalance in Predictive Maintenance for Machines using SMOTE-based Oversampling