A Temporal Convolutional Recurrent Neural Network Model for Epidemic Forecasting Using Multi-Dimensional Data: A new method (Preprint)

2020 ◽  
Author(s):  
Pu Miao ◽  
Xiaolong Zheng ◽  
Daniel Dajun Zeng
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Infectious Diseases ◽  
Recurrent Neural Network ◽  
Real World ◽  
Short Term Memory ◽  
Historical Data ◽  
Observation Data ◽  
Incidence Data ◽  
Epidemic Forecasting

BACKGROUND Infectious diseases such as COVID-19, influenza, Malaria, and Dengue have caused a significant threat throughout the world. For example, the expected yearly cost of pandemic influenza at roughly $500 billion, while COVID-19 has diminished the economic activity and could potentially lead to structural shifts in the global economy. One of the underlying major problems regarding these traditional surveillance epidemic methods is that they are not always effective and also the results produced by these methods usually have a delay of several weeks. OBJECTIVE The purpose of this study is to develop an epidemic forecasting model utilizing the deep learning technology that can be adapted to epidemic datasets and can predict the incidence or number of infectious diseases more accurately than traditional epidemic prediction methods. METHODS To predict the incidence of the epidemic, in this study, we collect real-world infectious disease data and transformed the dataset into time series. Our method uses the following information as inputs : (1) environmental and climatic information (2) epidemic–related internet search activity, (3) Google Trends, and (4) CDC ILI, Dengue Fever, Measles Incidence Data and related historical data. The proposed deep learning method utilizes a temporal convolutional technique that enables the exploitation of complex temporal patterns of epidemic activity across historical observations series. In the proposed deep learning model, we use the long and short-term memory units in a recurrent neural network to learn the temporal pattern of historical data. RESULTS We compare our model with three state-of-the-art deep learning models to evaluate the performance, accuracy, and relevance of the model predictions. We input the epidemic incidence data and observation data of the past 12 weeks to predict the number of incidence of the next one, two, and three weeks, respectively. We evaluate these models on the four real-world data sets we collected. The experiments demonstrate that our proposed model is better than several other models. CONCLUSIONS Previous studies often use autoregressive models or traditional machine learning methods to predict future epidemics. Compared to these, the performance evaluation of our method shows that our proposed method is superior to traditional non-machine methods and basic neural network models.

scholarly journals Comparative Analysis of Recurrent Neural Network Architectures for Reservoir Inflow Forecasting

Water ◽  
2020 ◽  
Vol 12 (5) ◽  
pp. 1500 ◽  
Author(s):  
Halit Apaydin ◽  
Hajar Feizi ◽  
Mohammad Taghi Sattari ◽  
Muslume Sevba Colak ◽  
Shahaboddin Shamshirband ◽  
...  
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Recurrent Neural Network ◽  
Energy Production ◽  
Short Term Memory ◽  
Efficiency Coefficient ◽  
Dam Reservoir ◽  
Short Term ◽  
Term Memory ◽  
Long Short Term Memory

Due to the stochastic nature and complexity of flow, as well as the existence of hydrological uncertainties, predicting streamflow in dam reservoirs, especially in semi-arid and arid areas, is essential for the optimal and timely use of surface water resources. In this research, daily streamflow to the Ermenek hydroelectric dam reservoir located in Turkey is simulated using deep recurrent neural network (RNN) architectures, including bidirectional long short-term memory (Bi-LSTM), gated recurrent unit (GRU), long short-term memory (LSTM), and simple recurrent neural networks (simple RNN). For this purpose, daily observational flow data are used during the period 2012–2018, and all models are coded in Python software programming language. Only delays of streamflow time series are used as the input of models. Then, based on the correlation coefficient (CC), mean absolute error (MAE), root mean square error (RMSE), and Nash–Sutcliffe efficiency coefficient (NS), results of deep-learning architectures are compared with one another and with an artificial neural network (ANN) with two hidden layers. Results indicate that the accuracy of deep-learning RNN methods are better and more accurate than ANN. Among methods used in deep learning, the LSTM method has the best accuracy, namely, the simulated streamflow to the dam reservoir with 90% accuracy in the training stage and 87% accuracy in the testing stage. However, the accuracies of ANN in training and testing stages are 86% and 85%, respectively. Considering that the Ermenek Dam is used for hydroelectric purposes and energy production, modeling inflow in the most realistic way may lead to an increase in energy production and income by optimizing water management. Hence, multi-percentage improvements can be extremely useful. According to results, deep-learning methods of RNNs can be used for estimating streamflow to the Ermenek Dam reservoir due to their accuracy.

scholarly journals A Statistics and Deep Learning Hybrid Method for Multivariate Time Series Forecasting and Mortality Modeling

Forecasting ◽  
2021 ◽  
Vol 4 (1) ◽  
pp. 1-25
Author(s):  
Thabang Mathonsi ◽  
Terence L. van Zyl
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Recurrent Neural Network ◽  
Short Term Memory ◽  
Hybrid Methods ◽  
Prediction Interval ◽  
Hybrid Approach ◽  
Computational Cost ◽  
Exponential Smoothing ◽  
Learning Methods

Hybrid methods have been shown to outperform pure statistical and pure deep learning methods at forecasting tasks and quantifying the associated uncertainty with those forecasts (prediction intervals). One example is Exponential Smoothing Recurrent Neural Network (ES-RNN), a hybrid between a statistical forecasting model and a recurrent neural network variant. ES-RNN achieves a 9.4% improvement in absolute error in the Makridakis-4 Forecasting Competition. This improvement and similar outperformance from other hybrid models have primarily been demonstrated only on univariate datasets. Difficulties with applying hybrid forecast methods to multivariate data include (i) the high computational cost involved in hyperparameter tuning for models that are not parsimonious, (ii) challenges associated with auto-correlation inherent in the data, as well as (iii) complex dependency (cross-correlation) between the covariates that may be hard to capture. This paper presents Multivariate Exponential Smoothing Long Short Term Memory (MES-LSTM), a generalized multivariate extension to ES-RNN, that overcomes these challenges. MES-LSTM utilizes a vectorized implementation. We test MES-LSTM on several aggregated coronavirus disease of 2019 (COVID-19) morbidity datasets and find our hybrid approach shows consistent, significant improvement over pure statistical and deep learning methods at forecast accuracy and prediction interval construction.

scholarly journals Hybrid deep learning model using recurrent neural network and gated recurrent unit for heart disease prediction

2021 ◽  
Vol 11 (6) ◽  
pp. 5467
Author(s):  
Surenthiran Krishnan ◽  
Pritheega Magalingam ◽  
Roslina Ibrahim
Keyword(s):  
Neural Network ◽  
Heart Disease ◽  
Deep Learning ◽  
Recurrent Neural Network ◽  
Short Term Memory ◽  
Learning Model ◽  
Disease Prediction ◽  
The Neural Network ◽  
Proposed Model ◽  
Deep Learning Model

<span>This paper proposes a new hybrid deep learning model for heart disease prediction using recurrent neural network (RNN) with the combination of multiple gated recurrent units (GRU), long short-term memory (LSTM) and Adam optimizer. This proposed model resulted in an outstanding accuracy of 98.6876% which is the highest in the existing model of RNN. The model was developed in Python 3.7 by integrating RNN in multiple GRU that operates in Keras and Tensorflow as the backend for deep learning process, supported by various Python libraries. The recent existing models using RNN have reached an accuracy of 98.23% and deep neural network (DNN) has reached 98.5%. The common drawbacks of the existing models are low accuracy due to the complex build-up of the neural network, high number of neurons with redundancy in the neural network model and imbalance datasets of Cleveland. Experiments were conducted with various customized model, where results showed that the proposed model using RNN and multiple GRU with synthetic minority oversampling technique (SMOTe) has reached the best performance level. This is the highest accuracy result for RNN using Cleveland datasets and much promising for making an early heart disease prediction for the patients.</span>

scholarly journals Research on the Estimate of Gas Hydrate Saturation Based on LSTM Recurrent Neural Network

Energies ◽  
2020 ◽  
Vol 13 (24) ◽  
pp. 6536
Author(s):  
Chuanhui Li ◽  
Xuewei Liu
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Recurrent Neural Network ◽  
Gas Hydrate ◽  
Short Term Memory ◽  
Chloride Concentration ◽  
Well Logs ◽  
Good Prediction ◽  
Learning Technology ◽  
Hydrate Saturation

Gas hydrate saturation is an important index for evaluating gas hydrate reservoirs, and well logs are an effective method for estimating gas hydrate saturation. To use well logs better to estimate gas hydrate saturation, and to establish the deep internal connections and laws of the data, we propose a method of using deep learning technology to estimate gas hydrate saturation from well logs. Considering that well logs have sequential characteristics, we used the long short-term memory (LSTM) recurrent neural network to predict the gas hydrate saturation from the well logs of two sites in the Shenhu area, South China Sea. By constructing an LSTM recurrent layer and two fully connected layers at one site, we used resistivity and acoustic velocity logs that were sensitive to gas hydrate as input. We used the gas hydrate saturation calculated by the chloride concentration of the pore water as output to train the LSTM network. We achieved a good training result. Applying the trained LSTM recurrent neural network to another site in the same area achieved good prediction of gas hydrate saturation, showing the unique advantages of deep learning technology in gas hydrate saturation estimation.

Combining Deep Learning Models for Enhancing the Detection of Botnet Attacks in Multiple Sensors Internet of Things Networks

2021 ◽  
Vol 5 (4) ◽  
pp. 380
Author(s):  
Abdulkareem A. Hezam ◽  
Salama A. Mostafa ◽  
Zirawani Baharum ◽  
Alde Alanda ◽  
Mohd Zaki Salikon
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Recurrent Neural Network ◽  
Short Term Memory ◽  
Denial Of Service ◽  
Learning Models ◽  
Short Term ◽  
Term Memory ◽  
Long Short Term Memory ◽  
Iot Devices

Distributed-Denial-of-Service impacts are undeniably significant, and because of the development of IoT devices, they are expected to continue to rise in the future. Even though many solutions have been developed to identify and prevent this assault, which is mainly targeted at IoT devices, the danger continues to exist and is now larger than ever. It is common practice to launch denial of service attacks in order to prevent legitimate requests from being completed. This is accomplished by swamping the targeted machines or resources with false requests in an attempt to overpower systems and prevent many or all legitimate requests from being completed. There have been many efforts to use machine learning to tackle puzzle-like middle-box problems and other Artificial Intelligence (AI) problems in the last few years. The modern botnets are so sophisticated that they may evolve daily, as in the case of the Mirai botnet, for example. This research presents a deep learning method based on a real-world dataset gathered by infecting nine Internet of Things devices with two of the most destructive DDoS botnets, Mirai and Bashlite, and then analyzing the results. This paper proposes the BiLSTM-CNN model that combines Bidirectional Long-Short Term Memory Recurrent Neural Network and Convolutional Neural Network (CNN). This model employs CNN for data processing and feature optimization, and the BiLSTM is used for classification. This model is evaluated by comparing its results with three standard deep learning models of CNN, Recurrent Neural Network (RNN), and long-Short Term Memory Recurrent Neural Network (LSTM–RNN). There is a huge need for more realistic datasets to fully test such models' capabilities, and where N-BaIoT comes, it also includes multi-device IoT data. The N-BaIoT dataset contains DDoS attacks with the two of the most used types of botnets: Bashlite and Mirai. The 10-fold cross-validation technique tests the four models. The obtained results show that the BiLSTM-CNN outperforms all other individual classifiers in every aspect in which it achieves an accuracy of 89.79% and an error rate of 0.1546 with a very high precision of 93.92% with an f1-score and recall of 85.73% and 89.11%, respectively. The RNN achieves the highest accuracy among the three individual models, with an accuracy of 89.77%, followed by LSTM, which achieves the second-highest accuracy of 89.71%. CNN, on the other hand, achieves the lowest accuracy among all classifiers of 89.50%.

scholarly journals Hand Gesture Recognition in Video Sequences Using Deep Convolutional and Recurrent Neural Networks

2020 ◽  
Vol 25 (1) ◽  
pp. 57-61
Author(s):  
Falah Obaid ◽  
Amin Babadi ◽  
Ahmad Yoosofan
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Learning ◽  
Gesture Recognition ◽  
Recurrent Neural Network ◽  
Short Term Memory ◽  
Hand Gesture ◽  
Short Term ◽  
Term Memory ◽  
Long Short Term Memory

AbstractDeep learning is a new branch of machine learning, which is widely used by researchers in a lot of artificial intelligence applications, including signal processing and computer vision. The present research investigates the use of deep learning to solve the hand gesture recognition (HGR) problem and proposes two models using deep learning architecture. The first model comprises a convolutional neural network (CNN) and a recurrent neural network with a long short-term memory (RNN-LSTM). The accuracy of model achieves up to 82 % when fed by colour channel, and 89 % when fed by depth channel. The second model comprises two parallel convolutional neural networks, which are merged by a merge layer, and a recurrent neural network with a long short-term memory fed by RGB-D. The accuracy of the latest model achieves up to 93 %.

scholarly journals Deep Learning Recurrent Neural Network for Concussion Classification in Adolescents Using Raw Electroencephalography Signals: Toward a Minimal Number of Sensors

2021 ◽  
Vol 15 ◽  
Author(s):  
Karun Thanjavur ◽  
Dionissios T. Hristopulos ◽  
Arif Babul ◽  
Kwang Moo Yi ◽  
Naznin Virji-Babul
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Learning ◽  
Recurrent Neural Network ◽  
Resting State ◽  
Short Term Memory ◽  
State Data ◽  
Eeg Data ◽  
Clinical Environments ◽  
Lstm Network

Artificial neural networks (ANNs) are showing increasing promise as decision support tools in medicine and particularly in neuroscience and neuroimaging. Recently, there has been increasing work on using neural networks to classify individuals with concussion using electroencephalography (EEG) data. However, to date the need for research grade equipment has limited the applications to clinical environments. We recently developed a deep learning long short-term memory (LSTM) based recurrent neural network to classify concussion using raw, resting state data using 64 EEG channels and achieved high accuracy in classifying concussion. Here, we report on our efforts to develop a clinically practical system using a minimal subset of EEG sensors. EEG data from 23 athletes who had suffered a sport-related concussion and 35 non-concussed, control athletes were used for this study. We tested and ranked each of the original 64 channels based on its contribution toward the concussion classification performed by the original LSTM network. The top scoring channels were used to train and test a network with the same architecture as the previously trained network. We found that with only six of the top scoring channels the classifier identified concussions with an accuracy of 94%. These results show that it is possible to classify concussion using raw, resting state data from a small number of EEG sensors, constituting a first step toward developing portable, easy to use EEG systems that can be used in a clinical setting.

scholarly journals Implementasi Metode Reccurrent Neural Network pada Text Summarization dengan Teknik Abstraktif

2019 ◽  
Vol 6 (4) ◽  
pp. 377
Author(s):  
Kasyfi Ivanedra ◽  
Metty Mustikasari
Keyword(s):  
Neural Network ◽  
Artificial Intelligence ◽  
Deep Learning ◽  
Recurrent Neural Network ◽  
Short Term Memory ◽  
Text Summarization ◽  
Short Term ◽  
Term Memory ◽  
Long Short Term Memory ◽  
F Measure

<p>Text Summarization atau peringkas text merupakan salah satu penerapan Artificial Intelligence (AI) dimana komputer dapat meringkas text pada suatu kalimat atau artikel menjadi lebih sederhana dengan tujuan untuk mempermudah manusia dalam mengambil kesimpulan dari artikel yang panjang tanpa harus membaca secara keseluruhan. Peringkasan teks secara otomatis dengan menggunakan teknik Abstraktif memiliki kemampuan meringkas teks lebih natural sebagaimana manusia meringkas dibandingkan dengan teknik ekstraktif yang hanya menyusun kalimat berdasarkan frekuensi kemunculan kata. Untuk dapat menghasilkan sistem peringkas teks dengan metode abstraktif, membutuhkan metode Recurrent Neural Network (RNN) yang memiliki sistematika perhitungan bobot secara berulang. RNN merupakan bagian dari Deep Learning dimana nilai akurasi yang dihasilkan dapat lebih baik dibandingkan dengan jaringan saraf tiruan sederhana karena bobot yang dihitung akan lebih akurat mendekati persamaan setiap kata. Jenis RNN yang digunakan adalah LSTM (Long Short Term Memory) untuk menutupi kekurangan pada RNN yang tidak dapat menyimpan memori untuk dipilah dan menambahkan mekanisme Attention agar setiap kata dapat lebih fokus pada konteks. Penelitian ini menguji performa sistem menggunakan Precision, Recall, dan F-Measure dengan membandingan hasil ringkasan yang dihasilkan oleh sistem dan ringkasan yang dibuat oleh manusia. Dataset yang digunakan adalah data artikel berita dengan jumlah total artikel sebanyak 4515 buah artikel. Pengujian dibagi berdasarkan data dengan menggunakan Stemming dan dengan teknik Non-stemming. Nilai rata-rata recall artikel berita non-stemming adalah sebesar 41%, precision sebesar 81%, dan F-measure sebesar 54,27%. Sedangkan nilai rata-rata recall artikel berita dengan teknik stemming sebesar 44%, precision sebesar 88%, dan F-measure sebesar 58,20 %.</p><p><em><strong>Abstract</strong></em></p><p class="Judul2"><em>Text Summarization is the application of Artificial Intelligence (AI) where the computer can summarize text of article to make it easier for humans to draw conclusions from long articles without having to read entirely. Abstractive techniques has ability to summarize the text more naturally as humans summarize. The summary results from abstractive techinques are more in context when compared to extractive techniques which only arrange sentences based on the frequency of occurrence of the word. To be able to produce a text summarization system with an abstractive techniques, it is required Deep Learning by using the Recurrent Neural Network (RNN) rather than simple Artificial Neural Network (ANN) method which has a systematic calculation of weight repeatedly in order to improve accuracy. The type of RNN used is LSTM (Long Short Term Memory) to cover the shortcomings of the RNN which cannot store memory to be sorted and add an Attention mechanism so that each word can focus more on the context.This study examines the performance of Precision, Recall, and F-Measure from the comparison of the summary results produced by the system and summaries made by humans. The dataset used is news article data with 4515 articles. Testing was divided based on data using Stemming and Non-stemming techniques.</em> <em>The average recall value of non-stemming news articles is 41%, precision is 81%, and F-measure is 54.27%. While the average value of recall of news articles with stemming technique is 44%, precision is 88%, and F-measure is 58.20%.</em></p><p><em><strong><br /></strong></em></p>

Sign in / Sign up

Export Citation Format

Share Document