LSTM Method Research Articles

Statistical downscaling of maximum temperature under CMIP6 global climate models and evaluation of heat wave events using deep learning methods for Indo‐Gangetic Plain

AbstractThe rising global temperature is one of the primary concerns of the world as it impacts the economy, environment and healthcare of any country which are more pronounced a regional level. Assessment of regional impacts of climate change at a local level requires fine resolution of climate data for which a robust and fast downscaling method is needed. In this study, we use three deep learning‐based methods, namely long short‐term memory network (LSTM), deep neural network (DNN) and recurrent neural network (RNN), to downscale CMIP6 13 GCMS models data (1.25° × 1.25° resolution) global climate model (GCM) maximum temperature (Tmax) at a regional scale of 0.5° × 0.5° spatial resolution for the period 1991–2010 over the Indo‐Gangetic Plain (IGP). In addition to the temperature prediction, heat wave events have been also analysed in the study. The study found that LSTM method performs better than DNN and RNN in downscaling of all GCM model datasets when evaluated against observed maximum temperature data from the India Meteorological Department (IMD) in terms of RMSE (0.9–3.5), average of all grid MAE value between (1.2 and 2.68), correlation (0.68–0.9) along with and spatiotemporal variability. LSTM also performed better in heat wave prediction over the region with similar temporal range (12–36 events) and spatial occurrence as compared to the observation (12–28 events). Overall, the study concludes that LSTM performs better than two methods for Indo‐Gangetic Plain with best hyper parameter tuning. Hence, we propose to utilize a deep learning framework based on LSTM for downscaling GCM dataset at a finer resolution.

Predicting the Energy Efficiency in Wireless Sensor Networks using LSTM and Random Forest Method

Predicting the Energy Efficiency in Wireless Sensor Networks using LSTM and Random Forest Method

Study on Univariate Modeling and Prediction Methods Using Monthly HIV Incidence and Mortality Cases in China.

AIDS presents serious harms to public health worldwide. In this paper, we used five single models: ARIMA, SARIMA, Prophet, BP neural network, and LSTM method to model and predict the number of monthly AIDS incidence cases and mortality cases in China. We have also proposed the LSTM-SARIMA combination model to enhance the accuracy of the prediction. This study provides strong data support for the prevention and treatment of AIDS. We collected data on monthly AIDS incidence cases and mortality cases in China from January 2010 to February 2024. Among them, for modeling, we used data from January 2010 to February 2021 and the rest for validation. Treatments were applied to the dataset based on its characteristics during modeling. All models in our study were performed using Python 3.11.6. Meanwhile, we used the constructed model to predict monthly incidence and mortality cases from March 2024 to July 2024. We then evaluated our prediction results using RMSE, MAE, MAPE, and SMAPE. The deep learning methods of LSTM and BPNN outperform ARIMA, SARIMA, and Prophet in predicting the number of mortality cases. When predicting the number of AIDS incidence cases, there is little difference between the two types of methods, and the LSTM method performs slightly better than the rest of the methods. Meanwhile, the average error in predicting AIDS mortality cases is significantly lower than in predicting AIDS incidence cases. The LSTM-SARIMA method outperforms other methods in predicting AIDS incidence and mortality. Due to the different characteristics of the AIDS incidence and mortality cases series, the performance of distinct methods is slightly different. The AIDS mortality series is smoother than the incidence series. The combined LSTM-SARIMA model outperforms the traditional method in prediction and the LSTM method alone, which is of practical significance for optimizing the prediction results of AIDS.

LSTM and Word Embedding: Classification and Prediction of Puskesmas Reviews Via Twitter

Puskesmas is a location for top-notch volunteer work that benefits the village and city governments alike. Therefore, patient feedback regarding the kinds of services offered by the community health center is required in an attempt to improve the quality service performance. Patient opinions can be expressed through reviews or opinions about the quality of patient care on social media sites like Facebook, Instagram, Twitter, WhatsApp, and Instagram. On the other hand, thoughts shared on social media are lengthy, unstructured texts. This complicates text analysis and makes it impossible to compare the caliber of services offered by Puskesmas managers. Furthermore, a number of Community Health Centers lack websites that allow users to rank Community Health Centers according to user interest and visual appeal and efficiency in operations. Thus, the purpose of this study is to classify and present sentiment analysis from Twitter about community health centers' health services. The scope focuses on five factors: administrative services, finances, mechanisms, health worker friendliness and skills, and administrative services. The LSTM word embedding model and the adadelta and adamax optimizers are used in word embedding for text mining. A confusion matrix was used to evaluate the developed model's degree of accuracy in categorizing and forecasting patient reviews. Results from the LSTM and Adamax models with a precision level of 76%, Recall 69% and Accuracy 71%. The results of this research show that the LSTM method and Adamax optimizer can classify and predict public opinion data about Puskesmas services via Twitter quite well. A high level of accuracy is very important to ensure that community opinions can be properly identified by the model, so that it can support the decision-making process in improving the type of Puskesmas services. To improve the model, further studies can be conducted on how to select parameters, select features, and create a quality dataset.

Extreme Weather Hazards and Insurance Revenues Based on Bi-directional LSTM

As extreme weather becomes more prevalent, it has serious implications for the insurance industry, real estate owners, and historic buildings. First of all, for the extreme weather, we summarize extreme weather into three parts: extreme temperature, extreme precipitation, and extreme humidity, and we process the data by combining the actual situation and applying the improved Critic method, and we demonstrate the data processing by taking the three data of Lhasa and Luxor as examples. anomalies. Subsequently, a future extreme climate prediction model for a region based on the Bi-directional LSTM method is developed. A time-weather model was fitted based on the collected data, with the highest and lowest points as the anomalies. At the same time, we simplify the formula to derive the insurance company's revenue, thus modeling the relationship between the degree of extreme weather hazard and insurance revenue, and derive that the insurance company will not be able to underwrite when the combined hazard is greater than the ratio of the insurance claim cost to the insurance payment. We applied the resulting model to Lhasa and Luxor, and plotted the combined hazard level of the actual data and the predicted data to help the insurance company visualize the combined hazard level of the region each month, and then choose whether to underwrite the policy in that region or not.

Comparative Analysis of Long Short-Term Memory Architecture for Text Classification

Text classification which is a part of NLP is a grouping of objects in the form of text based on certain characteristics that show similarities between one document and another. One of methods used in text classification is LSTM. The performance of the LSTM method itself is influenced by several things such as datasets, architecture, and tools used to classify text. On this occasion, researchers analyse the effect of the number of layers in the LSTM architecture on the performance generated by the LSTM method. This research uses IMDB movie reviews data with a total of 50,000 data. The data consists of positive, negative data and there is data that does not yet have a label. IMDB Movie Reviews data go through several stages as follows: Data collection, data pre-processing, conversion to numerical format, text embedding using the pre-trained word embedding model: Fastext, train and test classification model using LSTM, finally validate and test the model so that the results are obtained from the stages of this research. The results of this study show that the one-layer LSTM architecture has the best accuracy compared to two-layer and three-layer LSTM with training accuracy and testing accuracy of one-layer LSTM which are 0.856 and 0.867. While the training accuracy and testing accuracy on two-layer LSTM are 0.846 and 0.854, the training accuracy and testing accuracy on three layers are 0.848 and 864.

Power Consumption Predictive Analytics and Automatic Anomaly Detection Based on CNN-LSTM Neural Networks

In this modern era, electrical energy plays a crucial role in human life, as it is essential for most household appliances. The number of appliances requiring electrical energy increases each year, meeting the growing needs of users. However, electricity consumers tend to forget this fact and only realize its importance when they receive a significantly increased monthly electricity bill or face problems caused by anomalies in electricity use. Such anomalies can lead to substantial losses, especially when electrical equipment is damaged or left switched on without awareness. To make better decisions in such situations, real-time and accurate information is necessary, which can be achieved through data analytics utilizing machine-learning and predictive analytics. The purpose of this paper is to introduce the CNN-LSTM method of data analytic modeling for power consumption data collected through an electric data logger, which can help predict future power usage and detect real-time anomalies in the power network. The proposed model was tested using hourly electricity consumption data, and the results showed that the CNNLSTM method outperformed the LSTM model. The CNN-LSTM model had a 29% smaller Mean Squared Error (MSE) score than the LSTM method.

Application of Digital Health Concepts in Dance Culture Dissemination and Promotion Based on Predictive Modelling of Information Dissemination

Abstract The purpose of this paper is to analyze the dissemination effect of digital health concepts in dance culture by constructing a prediction model for dance culture information dissemination. It focuses on analyzing the main paths and forms of Internet+dance culture dissemination and constructing the cultural dissemination process of Internet+dance health. The LSTM method is used to extract the states in information dissemination, and the dissemination characteristics of digital health in dance culture are deduced by analyzing the characteristics of the information dissemination sequence. The association between digital health concepts and dance culture dissemination was determined through the trajectory characteristics of information dissemination. The results show that the spreading range of the four types of models rises in an upward trend and is faster when the spreading time round is in the interval of (0,125). The spread range in models in the (150,300) interval is generally higher than the SI model and SEI model. The factor value of whether the video is top or not affects the amount of playback is 0.771***, and the factor value of whether it carries a topic or not is 0.051.

Assurance of Network Communication Information Security Based on Cyber-Physical Fusion and Deep Learning

The internet brings high efficiency and convenience to society; however, the issue of information security in network communication has significantly affected every aspect of the society. How to ensure the security of this network communication information has become an important research topic. This paper proposes a diagnosis and prediction method based on cyber-physical fusion and deep learning, such as LSTM and CNN, to diagnose and predict network security in a complex network environment. The experiment results showed that the accuracy of network security diagnosis of the LSTM method in the training set was approximately 80%/ After the CNN training process, it has the highest accuracy rate of 95% on the test data set. This paper analysed the nature of network security problems from the perspective of cyber-physical fusion. CNN-based method to diagnose network security can obtain results with a higher accuracy rate so that technicians can better take measures to protect network security.

Offshore wind power output prediction based on convolutional attention mechanism

ABSTRACT To solve the problem that offshore wind power generation is susceptible to extreme weather conditions that result in low accuracy of monitoring data, this work proposes a method for predicting offshore wind power generation output by combining convolution and attention mechanisms. The method combines bidirectional long and short-term memory (BiLSTM) network with attention mechanism (AM). The input data are first weighted with the AM for reducing the predictive weight of the interference data, and then the attention mechanism-weighted data are fed into the BiLSTM network. The bi-directional propagation neural network can effectively utilize all the input information, resulting in higher prediction accuracy. The method of combining BiLSTM network with AM is compared with the method employing long and short-term memory network, gated recurrent unit, and bidirectional long and short-term memory network alone through simulation. The mean square error (MSE) of the BiLSTM combined with AM method is 41.28% smaller than the MSE of the best LSTM method among the compared methods and 58.43% smaller than the average of the compared methods. The R2-R-Square is 14.21% larger than the R2-R-Square of the LSTM and 32.91% larger than the average of the compared methods. The results show that the proposed method of combining convolution and attention mechanism for offshore wind power generation output prediction has higher prediction accuracy.

Biometric Personal Classification with Deep Learning Using EMG Signals

Biometric person recognition systems are becoming increasingly important due to their use in places requiring high security. Since it includes the physical and behavioral characteristics of people, the iris structure, which is a traditional person recognition system, is more secure than methods such as fingerprints or speech. In this study, a deep learning-based person classification/recognition model is proposed. The Gesture Recognition and Biometrics ElectroMyogram (GrabMyo) dataset from the open access PhysioNet database was used. With the 28-channel EMG device, 10 people were asked to make a fist movement with their hand. During the fist movement, data were recorded with the EMG device from the arm and wrist for 5 seconds with a sampling frequency of 2048. The EMD method was chosen to determine the spectral properties of EMG signals. With the EMD method, 4 IMF signal vectors were obtained from the high frequency components of the EMG signals. The classification performance effect of the feature vector is increased by using statistical methods for each IMF signal vector. Feature vectors are classified with CNN and LSTM methods from deep learning algorithms. Accuracy, Precision, Sensitivity and F-Score parameters were used to determine the performance of the developed model. An accuracy value of 95.57% was obtained in the model developed with the CNN method. In the LSTM method, the accuracy value was 93.88%. It is explained that the deep learning model proposed in this study can be effectively used as a biometric person recognition system for person recognition or classification problems with the EMG signals obtained during the fist movement. In addition, it is predicted that the proposed model can be used effectively in the design of future person recognition systems.

Prediksi Harga Forex Menggunakan Algoritma Long Short-Term Memory

Trillions of dollars per day of foreign currency trading activity occur in the forex market, which has very volatile movements in foreign currency trading. Trade based on bid and ask prices. The market determines foreign exchange rates based on supply and demand rules. Currency trading in pairs such as EUR/USD is a comparison of the value of the Euro against the Dollar as a basis for research, rising and falling currency prices in forex move fluctuatingly, so a market participant must be able to decide on buying and selling positions. Because wrong decisions can lead to losses. One of the ways to reduce risk in making decisions in buying and selling in forex can be using forecasting. This study uses the LSTM method in predicting forex prices which will be tested on several scales of dataset distribution. The smallest error results using a total dataset of 2631 with a dataset division of 70:15:15, which is divided into 70% data for training, 15% data for validation and 15% data for testing produces an RMSE value of 0.038, MAPE 2.5%. In measuring how well the regression model used with Rsquare on the data distribution is 70:15:15 and the total dataset used is 4979 to get the best results, namely 97%.

Prediksi Arah Harga Bitcoin Berdasarkan Manipulasi Metode Long Short-Term Memory (LSTM)

Bitcoin is one of the new currencies in online transactions that was found to be one of the cryptocurrencies that has fluctuating volatility, so researchers are looking for the determinants of bitcoin price. In this paper, we adopt a Long Short-Term Memory Networks (LSTM) approach to evaluate the effect of investor attention on bitcoin returns by constructing an aggregate investor attention proxy. In addition, LSTM is a method of Recurrent Neural Network (RNN) to deal with the problem of long-term dependence on RNN. The main objective of this research is to design, implement, and visualize Bitcoin Price Prediction using the LSTM method. The data this study are daily bitcoin cover data and tweets taken from twitter. The measure of the accuracy of the model used is the Root Mean Square Error (RMSE). The results show that the RMSE value is quite small, meaning that it is quite good in modeling bitcoin predictions.

A Novel Method of Fraud Detection of Credit Cards by Fuzzy, LSTM, and PSO Optimization

Since online shopping has become so popular, credit card theft has skyrocketed. This makes it hard to spot fake charges on accounts. In this research, credit card fraud detection is performed using a fuzzy database. It has been considered a data mining challenge to reliably identify whether or not a transaction is legitimate. This paper discusses the LSTM method and fuzzy logic. The learning process was sped up and made more precise by using a technique called particle swarm optimization (PSO). Performance values have been compared with those of the LSTM and fuzzy logic techniques, and a PSO-based neural network has been intensively trained by increasing the number of iterations and the population, or number of swarms. It has been shown that the PSO-based algorithm yields the best result for detecting fraudulent transactions. The goal of this method is to lower the mean square error (MSE) rate of the system. PSO is a popular optimization technique that can be used to locate the optimal set of features for the credit card fraud detection system. The proposed method PSO shows less mean squared error compared with Fuzzy and LSTM techniques.

A novel algorithm system for wind power prediction based on RANSAC data screening and Seq2Seq-Attention-BiGRU model

Accurate wind power prediction plays a crucial role in mitigating the challenges posed by the variability and fluctuations of wind power generation, while also assisting in the regulation of power grid peaks and voltages. This paper introduces a novel algorithm system for wind power prediction, which incorporates RANSAC noise screening and the Seq2Seq-Attention-BiGRU model to enhance prediction accuracy. To address the presence of large-scale noise datasets and the well-established correlation between wind speed and power, we propose the utilization of RANSAC for effective noise screening. Subsequently, the Seq2Seq-Attention wind power prediction model is developed and validated through comparisons with measured data from various wind farms across different seasons. Additionally, the BiGRU error correction model is introduced to further refine the wind power predictions. Comparative analyses between the proposed prediction method and existing forecasting techniques demonstrate the effectiveness of our approach in enhancing prediction accuracy. Specifically, when compared to the LSTM method, the RANSAC-Seq2Seq-Attention-BiGRU forecasting method yielded a remarkable decrease in the RMSE of 26.0%, 43.6%, 46.8%, and 48.4% for the months of January, March, June, and September, respectively. Moreover, the MAE exhibited reductions of 19.7%, 43.4%, 41.0%, and 46.2% for the same respective months.

Intelligent energy management strategy for hybrid electric vehicles using reinforcement learning

ABSTRACT Predicting the performance of a Hybrid EV (Electric Vehicle) is significant to alleviate the complications of insufficient fuel and other battery-related problems that affect the driving capacity. Many energy management techniques with ML (Machine Learning) and DL (Deep Learning) methodologies have been deployed in predicting the operation of vehicles based on energy sources and their features. Still, it is being noticed that prediction errors have increased along with the increase or decrease of input data. To solve the challenges addressed in existing studies, the proposed approach intends to use both ML and DL to predict faults from battery features that considerably manage the energy. It is accomplished through efficient feature extraction employed by Bi-LSTM (Bidirectional Long Short Term Memory), which extracts the best energy features required for predicting the defects. From the extracted input features collected hourly, daily, and monthly from the dataset, classification is done using the Modified XGBoost Algorithm with Empirical Loss function. Normal and Abnormal functioning of the Hybrid EV is classified effectively. Followed by that, the prediction of improper functioning is made with the assistance of the vehicle Id to maximise the vehicle’s performance through consideration of battery constraints. Experimentation is performed with MATLAB simulation, and the performance of the proposed system is evaluated using error metrics like RMSE (Root Mean Square Error), MAE (Mean Absolute Error), MSE (Mean Square Error), and accuracy. Efficiency is exhibited through the minimum loss obtained through the proposed work. From analysis of comparing LSTM method with proposed model, it is found that the LSTM method produced RMSE value of 1.055% which is 0.84% reduced in proposed model. On the other hand, the MAE of LSTM is 0.826% and the proposed model is decreased with 0.174%. Further, the MSE of LSTM is 1.113% and is 0.69% reduced in proposed model.

PREDIKSI HARGA PANGAN DI PASAR TRADISIONAL KOTA SURABAYA DENGAN METODE LSTM

Long Short-Term Memory is the development of an artificial neural network that has the ability to overcome the vanishing gradient problem, and makes it possible to remember long-term information, and understand temporal patterns in time series data, so that LSTM has good performance in predicting food prices [1]. In Indonesia, especially in Surabaya, food prices are often unstable. Fluctuations in food prices can be caused by many factors such as weather, growing season and production. Under these conditions, this research was conducted to predict future food prices. The purpose of this study is to apply the LSTM method in predicting food prices so that it can provide maximum results and can be used by the community in making good decisions. In this study the dataset used included 5 types of food, namely rice, beef, chicken eggs, granulated sugar, and cooking oil. The dataset was obtained from the website of the National Strategic Food Price Information Center (PIHPS Nasional, https://www.bi.go.id/hargapangan). Predictive results are evaluated with RMSE and MAE. RMSE and MAE values of 5 types of food, namely rice 32 and 27, beef 229 and 125, chicken eggs 319 and 213, cooking oil 424 and 215, granulated sugar 30 and 18.

Asian Stock Index Price Prediction Analysis Using Comparison of Split Data Training and Data Testing

This study implements stock index price predictions using the LSTM method, where one of the processes in data management before running with the LSTM method is data split. This study also looks for the most appropriate split data ratio in predicting stock index prices to minimize error rates and differences in forecasted prices and original prices because in previous studies there were several rules of thumb in dividing data, so it is necessary to compare the most appropriate ratios in this research. Based on the evaluation process, the error value was found from nine split data ratios that were run by five ratios which produced a predictive graph line shape that resembled the validation line. Three datasets, namely split data ratios of 80:20, 70:30, and 60:40, are the ratios that get the lowest error values based on the RMSE, MSE, MAPE, and MAE values in the five stock index datasets. The three ratios are then compared again by looking at the average percentage difference between the validation price and the predicted price for the next working day, and it is found that the ratio of 80:20 is the most suitable split data ratio for predicting the stock index price for the next working day, with a level of difference in the average value between the original price and the predicted price on the stock index of 1.3%. While the ratio of 70:30 has an average predicted value of five stock index datasets of 1.9% and a ratio of 60:40 of 1.8%.

Klasifikasi Opini Publik di Twitter Terhadap Bakal Calon Presiden Indonesia Tahun 2024 Menggunakan LSTM Secara Realtime Berbasis Website

The analysis of public opinions from Indonesian netizens regarding the potential presidential candidates for Indonesia in 2024 on Twitter is challenging. Human-based classification of the candidates on Twitter has limitations as it requires expertise and a considerable amount of time to process the data. Therefore, a system that provides realtime visualization of public opinion classification is necessary. Previous research only focused on model evaluation, while this study aims to implement the best model on a website. The objective of this research is to develop a system for monitoring the Twitter-based public opinion classification of the potential presidential candidates for Indonesia in 2024 within specific time frames. The training process utilizes the LSTM method, resulting in a model with an accuracy of 76%. Parameters such as batch size, dropout, and learning rate were tested. The data used in this study was obtained by crawling Twitter using the keywords Ganjar Pranowo, Anies Baswedan, and Prabowo Subianto. The LSTM model was then implemented in a website-based system that generates a dashboard with features such as a color-coded map displaying the highest levels of positive sentiment for each candidate in each province, the overall classification count for each candidate, and filters for sentiment classification based on province and specific time frames.

A Novel Approach for Plant Leaf Disease Predictions Using Recurrent Neural Network RNN Classification Method

Agriculture is an essential part of our country and is only necessary for its development. The problem here is leaf spot disease, which significantly affects the growth of food production. Food production is considered one of the most important in our nation. We must detect it early and bring it under control. It is important to analyze one of the versions of leaf disease in its development. The existing system of machine learning could not give the proper prediction result of the lead disease prediction and classification. Various plant diseases affect leaf areas, and leaf areas have some black marks. Black marks are one type of leaf disease in plants. Find out the disease levels and accuracy. One standard learning method of deep learning classification has been used to classify leaf disease prediction. Here pre-processing and feature extraction using CNN and feature extraction and type using the RNN-based LSTM method are used. The data has been found in the Kaggle dataset; the total dataset is divided into 20000 plant village dataset images using training, and the validation dataset has available there. This dataset predicts the disease name and related data is collected to gather the leaf disease information. The RNN algorithm result is improved accuracy of identification compared to other methods but best performance for the LSTM method.