Deep LSTM Research Articles

Influence of Artificial Intelligence on Forecasting Net Asset Value and Return Volatility in Indian Mutual Fund

Predicting the critical net asset value (NAV) in the financial market is difficult for investors and fund agencies. The present study introduces machine learning (ML) and deep learning (DL) models such as linear regression, deep long short-term memory recurrent neural network (DLSTM-RNN), and autoregressive integrated moving averages (ARIMA) for predicting the NAV. The five different equity sectoral technology mutual fund direct growth plans from January 2013 to December 2022 have been collected. The novelty of the current study is deeply examining, which ML or DL model devotedly predicts the NAV closing price. The major key findings of the experimental results proved that the DLSTM-RNN model makes statistically viable predictions, whereby the mean absolute percent error (MAPE) average prediction accuracy value is 0.02. Based on the accuracy of a superior model, we compute the annualized return volatility to compare the risk of investments with annual return periods over different time horizons. The Jarque-Bera statistics of the return volatility over time Gaussian distribution is rejected at the 0.01 level. Statistical paired t-test and Pearson correlation coefficient are used to compare the effects of the proposed three models. In addition, the benchmark portfolio strategy yields a Sharpe ratio of 7.0193 and the maximum drawdown is 0.3743. The AI performed deep LSTM neural network model simulation, especially when using a daily and monthly MAPE strategy giving 81% and 84% highest NAV prediction consistency than the linear regression and ARIMA models.

Whale-optimized LSTM networks for enhanced automatic text summarization.

Automatic text summarization is a cornerstone of natural language processing, yet existing methods often struggle to maintain contextual integrity and capture nuanced sentence relationships. Introducing the Optimized Auto Encoded Long Short-Term Memory Network (OAELSTM), enhanced by the Whale Optimization Algorithm (WOA), offers a novel approach to this challenge. Existing summarization models frequently produce summaries that are either too generic or disjointed, failing to preserve the essential content. The OAELSTM model, integrating deep LSTM layers and autoencoder mechanisms, focuses on extracting key phrases and concepts, ensuring that summaries are both informative and coherent. WOA fine-tunes the model's parameters, enhancing its precision and efficiency. Evaluation on datasets like CNN/Daily Mail and Gigaword demonstrates the model's superiority over existing approaches. It achieves a ROUGE Score of 0.456, an accuracy rate of 84.47%, and a specificity score of 0.3244, all within an efficient processing time of 4,341.95 s.

Forest Change Detection and Prediction Using the Crow Chicken Swarm Optimization-Based Deep LSTM Classifier

Forest Change Detection and Prediction Using the Crow Chicken Swarm Optimization-Based Deep LSTM Classifier

Optimizing the LSTM Deep Learning Model for Arctic Sea Ice Melting Prediction

Optimizing the LSTM Deep Learning Model for Arctic Sea Ice Melting Prediction

A Modelling Framework of an Intelligent News Recommendation Engine for Financial Analysts

Referring to news articles has become a predominant task in the daily routine of each financial analyst, as it helps to make accurate financial insights. However, as there are thousands of news articles generated daily, finding the most relevant articles becomes a time-consuming task. Hence, this study develops a modelling framework of an intelligent news recommendation engine for financial analysts in a particular financial company, which assists in recommending the most appropriate articles according to analysts' preferences in an efficient and effective manner without tedious browsing.In this study, the data collection phase of the recommendation engine is accomplished by retrieving news articles from online news websites using the web scraping technique. The response variable, analysts’ preference level for each article is manually acquired from a group of financial analysts at a selected financial company. In the analysis phase of the recommendation engine, a classification-based method was utilized where three machine learning models, KNN, SVM, Random Forest, and two deep learning models, LSTM and CNN with Natural Language Processing techniques are experimented to discover the best algorithm. Moreover, the synonym replacement method is employed as a text data augmentation method to address the imbalanced problem in the dataset.As CNN obtains the highest accuracy in comparison to other applied methods it is chosen as the most suitable approach in the analysis phase. Moreover, this study reveals that DL models perform considerably higher performances in the context of news recommendation engines rather than ML approaches. Lastly, this framework can be adjustable for any financial company with minor modifications.

Developing An Interactive Web-Based Time Series Forecasting System With Deep Learning And LSTM For Student Enrollment Using Dash

Developing An Interactive Web-Based Time Series Forecasting System With Deep Learning And LSTM For Student Enrollment Using Dash

Deep Learning Autoencoder Study on ECG Signals

Arrhythmia refers to an irregular heart rhythm resulting from disruptions in the heart's electrical activity. To identify arrhythmias, an electrocardiogram (ECG) is commonly employed, as it can record the heart's electrical signals. However, ECGs may encounter interference from sources like electromagnetic waves and electrode motion. Several researchers have investigated the denoising of electrocardiogram signals for arrhythmia detection using deep autoencoder models. Unfortunately, these studies have yielded suboptimal results, indicated by low Signal-to-Noise Ratio (SNR) values and relatively large Root Mean Square Error (RMSE). This study addresses these limitations by proposing the utilization of a Deep LSTM Autoencoder to effectively denoise ECG signals for arrhythmia detection. The model's denoising performance is evaluated based on achieved SNR and RMSE values. The results of the denoising evaluations using the Deep LSTM Autoencoder on the AFDB dataset show SNR and RMSE values of 56.16 and 0.00037, respectively. Meanwhile, for the MITDB dataset, the corresponding values are 65.22 and 0.00018. These findings demonstrate significant improvement compared to previous research. However, it's important to note a limitation in this study—the restricted availability of arrhythmia datasets from MITDB and AFDB. Future researchers are encouraged to explore and acquire a more extensive collection of arrhythmia data to further enhance denoising performance.

COVID-19 prediction using Caviar Squirrel Jellyfish Search Optimization technique in fog-cloud based architecture.

In the pandemic of COVID-19 patients approach to the hospital for prescription, yet due to extreme line up the patient gets treatment after waiting for more than one hour. Generally, wearable devices directly measure the preliminary data of the patient stored in capturing mode. In order to store the data, the hospitals require large storage devices that make the progression of data more complex. To bridge this gap, a potent scheme is established for COVID-19 prediction based fog-cloud named Caviar Squirrel Jellyfish Search Optimization (CSJSO). Here, CSJSO is the amalgamation of CAViar Squirrel Search Algorithm (CSSA) and Jellyfish Search Optimization (JSO), where CSSA is blended by the Conditional Autoregressive Value-at-Risk (CAViar) and Squirrel Search Algorithm (SSA). This architecture comprises the healthcare IoT sensor layer, fog layer and cloud layer. In the healthcare IoT sensor layer, the routing process with the collection of patient health condition data is carried out. On the other hand, in the fog layer COVID-19 detection is performed by employing a Deep Neuro Fuzzy Network (DNFN) trained by the proposed Remora Namib Beetle JSO (RNBJSO). Here, RNBJSO is the combination of Namib Beetle Optimization (NBO), Remora Optimization Algorithm (ROA) and Jellyfish Search optimization (JSO). Finally, in the cloud layer, the detection of COVID-19 employing Deep Long Short Term Memory (Deep LSTM) trained utilizing proposed CSJSO is performed. The evaluation measures utilized for CSJSO_Deep LSTM in database-1, such as Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) observed 0.062 and 0.252 in confirmed cases. The measures employed in database-2 are accuracy, sensitivity and specificity achieved 0.925, 0.928 and 0.925 in K-set.

HiTrip: Historical trajectory interpolation for trawlers via deep learning on multi-source data

The Vessel Monitoring System (VMS) on trawlers has revolutionized our understanding of spatiotemporal fishing activities. However, the low temporal resolution of historical VMS datasets complicates a precise analysis of fishing effort distribution. One inherent challenge for precise interpolation is the stark contrast between trawler movement patterns during steaming, characterized by straight lines, and fishing, which often involves consecutive turns. In this study, we introduce HiTrip, a deep learning approach that interpolates historical VMS data from two-hour intervals down to three minutes by harnessing both VMS and marine hydrological datasets. The proposed deep learning model, integrating ResNet, LSTM, and MLP, seamlessly synthesizes spatial features from coarse fishing effort distributions, sea surface factor fields, and current fields, while accounting for the temporal relationships within trajectory segments. Evaluated on 1855 East China Sea trawler VMS records and Copernicus Climate Data Store hydrological factor data, HiTrip achieves a 0.20 km interpolation error, meeting a finery 0.005° × 0.005°spatial resolution demand for fishing effort distribution analysis. Ablation study validates the efficacy of our deep learning model integrating multi-source datasets. Moreover, when evaluated on a diverse Global Fishing Watch dataset, which includes 45 trawlers spanning various global maritime regions, HiTrip maintains a 0.40 km error, emphasizing its broad generalization ability.

Short-term forecasting airport passenger flow during periods of volatility: Comparative investigation of time series vs. neural network models

Recurrent Neural Networks (RNNs), known for handling complex data tasks like language translation and speech recognition, are seldom employed in airport management practice for daily and weekly passenger flow forecasting tasks. In this paper, we evaluate the effectiveness and adaptability of various neural network models (RNN, LSTM, GRU, Deep LSTM, Bidirectional LSTM, multivariate RNN, and multivariate LSTM) against standard time series models (ARIMA, SARIMA, and SARIMAX) for a short-term forecasting airport security checkpoint passenger flows at five major U.S. airports during the pandemic.At Atlanta’s Hartsfield-Jackson Airport (ATL), the RNN notably surpasses SARIMA’s forecasting accuracy by 34% (DM = 3.44, p < 0.01). This underscores RNN’s superiority in handling complex interactions among variables and non-linear dynamics, demonstrating its readiness for the emerging data-rich environment. Including exogenous variables enhances the forecasting accuracies of the multivariate RNN/LSTM (DM = 6.82, p < 0.01; DM = 2.65, p < 0.01, respectively), while the SARIMAX struggles with the added complexity.We observed the same patterns at the other four airports studied (DEN/ORD/LAX/DFW) during the pandemic period. However, during the normal airport traffic period, the clear superiority of RNN became much less pronounced, obscuring the performance gap between RNN and SARIMA. This suggests that the inherent advantages of RNN in capturing non-linearity are accentuated during volatile conditions and less pronounced or not pronounced at all during routine periods.

Adam Adadelta Optimization based bidirectional encoder representations from transformers model for fake news detection on social media

Social platform have disseminated the news in rapid speed and has been considered an important news resource for many people over worldwide because of easy access and less cost benefits when compared with the traditional news organizations. Fake news is the news deliberately written by bad writers that manipulates the original contents and this rapid dissemination of fake news may mislead the people in the society. As a result, it is critical to investigate the veracity of the data leaked via social media platforms. Even so, the reliability of information reported via this platform is still doubtful and remains a significant obstacle. As a result, this study proposes a promising technique for identifying fake information in social media called Adam Adadelta Optimization based Deep Long Short-Term Memory (Deep LSTM). The tokenization operation in this case is carried out with the Bidirectional Encoder Representations from Transformers (BERT) approach. The measurement of the features is reduced with the assistance of Kernel Linear Discriminant Analysis (LDA), and Singular Value Decomposition (SVD) and the top-N attributes are chosen by employing Renyi joint entropy. Furthermore, the LSTM is applied to identify false information in social media, with Adam Adadelta Optimization, which comprises a combo of Adam Optimization and Adadelta Optimization . The Deep LSTM based on Adam Adadelta Optimization achieved maximum accuracy, sensitivity, specificity of 0.936, 0.942, and 0.925.

Financial Risk Prediction Model in the Context of Big Data - Corporate Financial Risk Control Based on LSTM Deep Neural Networks

Abstract This paper is based on the use of recurrent neural networks and LSTM deep neural networks to obtain the financial risk prediction feature sequence in the context of big data. The financial risk prediction feature sequence is used as the input value of the input gate of the LSTM deep neural network model after data filtering, normalization and loss function optimization, and then the financial risk prediction for the output gate of the LSTM deep neural network model. Considering the availability of data, small and medium-sized enterprises listed in A-share companies in the Wind database are selected as sample enterprises, and evaluation indexes are constructed and detected at the same time so as to complete the experimental design of enterprise financial risk prediction in the context of big data. The prediction of enterprise financial risk is empirically analyzed using simulation analysis and statistical analysis. The results show that in the model performance analysis, the average value of ten years of data, the highest value is still the result obtained by LSTM training, 0.761, compared with other models of LSTM deep neural network in static financial risk prediction in the overall best performance. In the case study of Yibai Pharmaceutical, the minimum value of the rate of return, return on total assets, and return on assets were -10.02%, 2.56%, -20.72%, which reflects the fact that the private enterprises still have large profitability space to be mined. This study helps investors or financial institutions such as funds to find out the possible financial risk crisis of listed companies as early as possible to avoid the parties from incurring large financial losses.

Portfolio Optimization Based on Deep Learning and Factor Constraints

Portfolio construction can help investors achieve a balance between risks and returns, and rationally allocate funds to maximize returns. This paper combines four deep learning models LSTM, GRU, CRNN, TCN with investment portfolio strategies Mean-Variance, Mean-CVaR, selects stock data from different industries in the US stock market to construct portfolios, and these strategies are tested in both bull and bear market environment. Comparative analysis of cumulative returns reveals that in the bull market, the cumulative return of TCN+MV and GRU+MV is the highest. In the bear market, the cumulative returns of portfolios using the four deep learning algorithms combined with MV are similar and overall outperform those combined with MC. Furthermore, based on the deep learning algorithm and MV model, this paper selects multiple factors for scoring, and uses factor scores as constraints to the process of portfolio optimization, and the cumulative return has been significantly improved. The method in this paper can provide a theoretical reference for investors to construct investment portfolios and weigh risks and benefits according to individual needs.

A novel Deep Reinforcement Learning based automated stock trading system using cascaded LSTM networks

Deep Reinforcement Learning (DRL) algorithms have been increasingly used to construct stock trading strategies, but they often face performance challenges when applied to financial data with low signal-to-noise ratios and unevenness, as these methods were originally designed for the gaming community. To address this issue, we propose a DRL-based stock trading system that leverages Cascaded Long Short-Term Memory (CLSTM-PPO Model) to capture the hidden information in the daily stock data. Our model adopts a cascaded structure with two stages of carefully designed deep LSTM networks: it uses one LSTM to extract the time-series features from a sequence of daily stock data in the first stage, and then the features extracted are fed to the agent in the reinforcement learning algorithm for training, while the actor and the critic in the agent also use a LSTM network. We conduct experiments on stock market datasets from four major indices: the Dow Jones Industrial index (DJI) in the US, the Shanghai Stock Exchange 50 (SSE50) in China, S&P BSE Sensex Index (SENSEX) in India, and the Financial Times Stock Exchange 100 (FTSE100) in the UK. We compare our model with several benchmark models, including: (i) a model based on a buy-and-hold strategy; (ii) a Proximal Policy Optimization (PPO) model with Multilayer Perceptron (MLP) policy; (iii) some up-to-date models like the MLP model, LSTM model, Light Gradient Boosting Machine (LGBM) model, and histogram-based gradient boosting model; and (iv) an ensemble strategy model. The experimental results show that our model outperforms the baseline models in several key metrics, such as cumulative returns, maximum earning rate, and average profitability per trade. The improvements range from 5% to 52%, depending on the metric and the stock index. This indicates that our proposed method is a promising way to build an automated stock trading system.

Employing LSTMs for Music Remixing

Abstract: The Main Idea was to generate a mixed music from solo musical instrument. For carrying out this task there are various subtask to be carried from data collection to preprocessing the data to make it compatible with the Long Short Term Memory Recurrent Neural Network (LSTM RNNs) model. So, Music mp3 media was collected from melody loops and each one was trimmed down to 30 seconds. Conversion tool was used for segregating the Music into four different mp3 media file i.e. Bass.mp3, Drums.mp3, piano.mp3 and others.mp3. We prepared different models for accomplishing the task to get good results; therefore we preprocessed and transformed the data every time for the new Model architecture. We started with developing the most basic model architecture using Deep neural network (DNNs) and LSTM RNN, and this model architecture was so basic that it did not captures the dependencies within the music , although results was fine but music it generated does not make any sense. Than we tried experimenting by developing different model architectures which generates sensible music and we developed more complex architectures using LSTM RNNs which have the potential to make good results. But there is one disadvantage of it that it takes very long time to get fully trained. In order to reduce the training time and utilizing our resources properly we divided our model into 4 sub-models each for generating music of different musical instruments , These sub-models were trained in parallel on different Google Colaboratory notebook and their best model weights were stored for Reusing them for inference. These all were put into single notebook for predictions and each of these predictions are then combined to make the desired result

Human Activity Recognition Using CNN and Lstm Deep Learning Algorithms

Human Activity Recognition recognizes and classifies the activities performed by the users or people based on the data collected from the sensors of special devices such as smart-watches, smartphones etc. It has become easy to collect a huge amount of data from inertial sensors that are embedded in wearable devices. An accelerometer and gyroscope sensors are most commonly used inertial sensors. There are various already available datasets, in our paper, we are using the Wireless Sensor Data Mining dataset which contains 1,098,207 data of 6 physical activities performed. In this paper, the activities we aim to classify are walking, jogging, going up and downstairs, standing, and sitting. There are various algorithms applied on the various datasets. In our paper, we use Convolutional Neural Network and Long Short-Term Memory deep learning algorithm on the data set, we split the data into training data [80%] and testing data [20%]. By using a confusion matrix, we recognize and classify the activities performed using maximum accuracy.

Assessing the impact of borehole field data on AI-based deep learning models for heating and cooling prediction

Sensor readings play a critical role in the prediction performance of AI-based data-driven fault-detection and diagnostics for borehole heat exchangers in heating and cooling systems. The accuracy and algorithm selection of individual deep neural networks are well-investigated research topics but the impact of quality of sensors data on algorithm selection and its further impact on prediction performance has not been investigated and quantified for geothermal heating/cooling systems. The objective of the current work is to investigate the impact of data quality, available as measured time-series data from the borehole field, on the prediction accuracy of the deep neural networks-based models. In the current work four different data sets (D-1, D-2, D-3, D-4) are used for data-driven modeling using four different deep machine learning approaches, i.e., LSTM, BD-LSTM, GRU and CNN. The developed AI models have been trained, validated, and tested using the real-field data sets with same number of independent input parameters i.e., inlet temperature and mass flow rate to predict the fluid outlet temperature. The validated AI models are used to predict the performance of borehole heat exchanger to meet the heating and cooling loads. The prediction results show that the BD-LSTM outperforms other deep neural networks in terms of accuracy. With the data sets D-1, D-2, D-3 and D-4 it was observed that BD-LSTM had lowest MAPE of 2.99 %, 0.22 %, 6.39 % and 2.21 % respectively. Several experiments show that the BD-LSTM and LSTM models are better than the other used approaches in terms of prediction accuracy, but larger number of trainable parameters results in more computation time than CNN and GRU respectively. More data points with higher sampling frequency, results in more accurate prediction as depicted by the results from D-2 and D-4. Although, the results showed that the data collection interval is crucial, but different test runs indicate that avoiding the loss of time-series data during the cleaning process is equally important to accurately capture the Seasonality and Trend of the heating/cooling data. This is realized in case of D-1 where almost 58 % of the values were removed as outliers and these missing values were needed to be approximated which also affected the model prediction accuracy. The presented analysis highlights that the LSTM based DNN model has the capability to predict borehole heat exchanger performance accurately and the predictions are highly dependent on data frequency, the quality as well as data quantity from geothermal installations used to train the model.

Design of a Soft Sensor Based on Long Short-Term Memory Artificial Neural Network (LSTM) for Wastewater Treatment Plants.

Assessment of wastewater effluent quality in terms of physicochemical and microbial parameters is a difficult task; therefore, an online method which combines the variables and represents a final value as the quality index could be used as a useful management tool for decision makers. However, conventional measurement methods often have limitations, such as time-consuming processes and high associated costs, which hinder efficient and practical monitoring. Therefore, this study presents an approach that underscores the importance of using both short- and long-term memory networks (LSTM) to enhance monitoring capabilities within wastewater treatment plants (WWTPs). The use of LSTM networks for soft sensor design is presented as a promising solution for accurate variable estimation to quantify effluent quality using the total chemical oxygen demand (TCOD) quality index. For the realization of this work, we first generated a dataset that describes the behavior of the activated sludge system in discrete time. Then, we developed a deep LSTM network structure as a basis for formulating the LSTM-based soft sensor model. The results demonstrate that this structure produces high-precision predictions for the concentrations of soluble X1 and solid X2 substrates in the wastewater treatment system. After hyperparameter optimization, the predictive capacity of the proposed model is optimized, with average values of performance metrics, mean square error (MSE), coefficient of determination (R2), and mean absolute percentage error (MAPE), of 23.38, 0.97, and 1.31 for X1, and 9.74, 0.93, and 1.89 for X2, respectively. According to the results, the proposed LSTM-based soft sensor can be a valuable tool for determining effluent quality index in wastewater treatment systems.

Ultra-short-term PV power prediction based on Informer with multi-head probability sparse self-attentiveness mechanism

As a clean energy source, solar power plays an important role in reducing the high carbon emissions of China’s electricity system. However, the intermittent nature of the system limits the effective use of photovoltaic power generation. This paper addresses the problem of low accuracy of ultra-short-term prediction of distributed PV power, compares various deep learning models, and innovatively selects the Informer model with multi-head probability sparse self-attention mechanism for prediction. The results show that the CEEMDAN-Informer model proposed in this paper has better prediction accuracy, and the error index is improved by 30.88% on average compared with the single Informer model; the Informer model is superior to other deep learning models LSTM and RNN models in medium series prediction, and its prediction accuracy is significantly better than the two. The power prediction model proposed in this study improves the accuracy of PV ultra-short-term power prediction and proves the feasibility and superiority of the deep learning model in PV power prediction. Meanwhile, the results of this study can provide some reference for the power prediction of other renewable energy sources, such as wind power.

Developing Information Gateway for Intelligent Decision-Making and Stability in Solar Power Generation

Abstract This study has developed a specialized information gateway for inverters, utilizing relays, data converters, and single-board computers, among other components. Upon receiving data from the gateway, the server processes it to generate an intelligent solar monitoring system. The platform utilizes deep learning RNN and LSTM algorithms to forecast power generation at solar power plants, allowing for real-time monitoring of weather and power generation. By comparing actual and expected power generation data, the system can adjust equipment maintenance and cleaning schedules. It is designed to automatically alert staff members to take appropriate action when anomalous power generation data is continuously transmitted. Additionally, the system sends an alert for on-site inspection and removal of any abnormal situations to increase the stability of solar power generation. This study employs deep learning and IoT data collection to provide the knowledge necessary for intelligent decision-making and increased stability in solar power generation. JEL classification numbers: C43, F68, H41. Keywords: Solar energy, IoT monitoring, RNN, LSTN, Abnormal return.