Long-term Dependencies Research Articles

Enhancing agricultural product trade efficiency through machine learning predictions and multi-objective optimization of financial strategies

In the context of global economic development and food safety, agricultural trade plays a vital role in linking agricultural production with market demand, and the efficient formulation of agricultural financial strategies is crucial for enhancing trade efficiency. This study employs advanced machine learning technologies to predict and optimize the efficiency of agricultural trade while seeking to balance the interests of various stakeholders in agricultural finance. By integrating quantile factor models, long short-term memory (LSTM) networks, and attention mechanisms, this paper conducts an in-depth analysis and precise prediction of the key factors affecting trade efficiency. This approach effectively addresses the nonlinearity and long-term dependency issues in time-series data, utilizing attention mechanisms to highlight critical information and improve prediction accuracy. Furthermore, this research establishes a multi-objective optimization model to balance the interests of agricultural finance participants, providing a new quantitative tool for formulating agricultural financial strategies aimed at optimizing decisions to enhance economic and social value. This paper offers new perspectives, methods, and empirical support for improving the efficiency of agricultural trade and the formulation of agricultural financial strategies.

A compound framework for short-term gas load forecasting combining time-enhanced perception transformer and two-stage feature extraction

The development of accurate and reliable gas load forecasting models is critical to residential safety and gas company economics. However, since the seq2seq neural network conveys information by sequentially updating the hidden states, when processing gas load data, the information needs to be passed several times before it can affect the prediction results at subsequent time steps. This can lead to a lack of ability of the model to capture long-term dependencies. To address the above issues, we propose a novel compound framework. Time-enhanced perception transformer (TEPT) in the framework achieves parallel computation of historical data and stronger location awareness through convolutional self-attention and improved attention scores. This results in better capture of local patterns and long-term dependencies in gas loads. In addition, the framework proposes two-stage feature extraction (TSFE) for decomposing the important parts of the original information and extracting more refined features based on them. We constructed a real gas load data set from Wuxi, China to verify the effectiveness. The results show that MAE, MAPE, and RMSE of the proposed method are 3.45, 1.07%, and 4.68, which provide higher accuracy compared to other prediction methods.

Multi scale encoder-decoder network with Time Frequency Attention and S-TCN for single channel speech enhancement

The goal of speech enhancement is to restore clean speech in noisy environments. Acoustic scenarios with low signal-to-noise ratios (SNR) make it quite challenging to extract the target speech from its noise. In the current study, to enhance noisy speech, we propose a feature recalibration based multi-scale convolutional encoder-decoder architecture with squeeze temporal convolutional networks (S-TCN) bottleneck. Each multi-scale convolutional layer in encoder and decoder is followed by time-frequency attention module (TFA). The recalibration based multi-scale 2D convolution layers are used to extract local and contextual information. Additionally, the recalibration network is equipped with a gating mechanism to control the flow of information among the layers, enabling weighting of the scaled features for noise suppression and speech retention. The fully connected layer (FC) in the bottleneck part of encoder-decoder contains a few neurons, which capture the global information from the multi-scale 2D convolution layer and reduce parameters. A S-TCN, inspired by the popular temporal convolutional neural network (TCNN), is inserted between the encoder and the decoder to model long-term dependencies in speech. The TFA is a highly efficient network component, that operates through two simultaneous attentions, one focused on time frames, and the other on frequency channels. These attentions work together to explicitly exploit positional information to create a two-dimensional attention map to effectively capture the significant time-frequency distribution of speech. Utilizing the common voice dataset, our proposed model consistently enhances results compared to the current benchmarks, as demonstrated by two extensively utilized objective measures PESQ and STOI. The proposed model shows significant improvements, with average PESQ and STOI scores increasing by 45.7% and 23.8% respectively for seen background noises, and by 43.5% and 21.4% for unseen background noises, when compared to the quality of noisy speech. Tests validate that the proposed approach outperforms numerous cutting-edge algorithms.

Data-Driven 4D Trajectory Prediction Model Using Attention-TCN-GRU

With reference to the trajectory-based operation (TBO) requirements proposed by the International Civil Aviation Organization (ICAO), this paper concentrates on the study of four-dimensional trajectory (4D Trajectory) prediction technology in busy terminal airspace, proposing a data-driven 4D trajectory prediction model. Initially, we propose a Spatial Gap Fill (Spat Fill) method to reconstruct each aircraft’s trajectory, resulting in a consistent time interval, noise-free, high-quality trajectory dataset. Subsequently, we design a hybrid neural network based on the seq2seq model, named Attention-TCN-GRU. This consists of an encoding section for extracting features from the data of historical trajectories, an attention module for obtaining the multilevel periodicity in the flight history trajectories, and a decoding section for recursively generating the predicted trajectory sequences, using the output of the coding part as the initial input. The proposed model can effectively capture long-term and short-term dependencies and repetitiveness between trajectories, enhancing the accuracy of 4D trajectory predictions. We utilize a real ADS-B trajectory dataset from the airspace of a busy terminal for validation. The experimental results indicate that the data-driven 4D trajectory prediction model introduced in this study achieves higher predictive accuracy, outperforming some of the current data-driven trajectory prediction methods.

MEAformer: An all-MLP transformer with temporal external attention for long-term time series forecasting

Transformer-based models have significantly improved performance in Long-term Time Series Forecasting (LTSF). These models employ various self-attention mechanisms to discover long-term dependencies. However, the computational efficiency is hampered by the inherent permutation invariance of self-attention, and they primarily focus on relationships within the sequence while neglecting potential relationships between different sample sequences. This limits the ability and flexibility of self-attention in LTSF. In addition, the Transformer's decoder outputs sequences in an autoregressive manner, leading to slow inference speed and error accumulation effects, especially for LTSF. Regarding the issues with Transformer-based models for LTSF, we propose a model better suited for LTSF, named MEAformer. MEAformer adopts a fully connected Multi-Layer Perceptron (MLP) architecture consisting of two types of layers: encoder layers and MLP layers. Unlike most encoder layers in Transformer-based models, the MEAformer replaces self-attention with temporal external attention. Temporal external attention explores potential relationships between different sample sequences in the training dataset. Compared to the quadratic complexity of self-attention mechanisms, temporal external attention has efficient linear complexity. Encoder layers can be stacked multiple times to capture time-dependent relationships at different scales. Furthermore, the MEAformer replaces the intricate decoder layers of the original model with more straightforward MLP layers. This modification aims to enhance inference speed and facilitate single-pass sequence generation, effectively mitigating the problem of error accumulation effects. Regarding long-term forecasting, MEAformer achieves state-of-the-art performance on six benchmark datasets, covering five real-world domains: energy, transportation, economy, weather, and disease. Code is available at: https://github.com/huangsiyuan924/MEAformer.

Anomaly Detection for Landslide Displacement Monitoring Data Based on TCN-Transformer

In view of the high false alarm rate of landslide early warning, this paper proposes a classification method of landslide displacement monitoring data based on TCN and attention mechanism to identify normal data and abnormal data. First, the original monitoring data is processed and denoised to improve the data quality. Then, TCN is used to capture the long-term dependence of time series data and highlight key features in combination with attention mechanism. Finally, the landslide displacement monitoring data can be accurately classified by building a classification model to find out the abnormal data and reduce the false alarm rate of early warning. Experiments show that this method can recognize abnormal displacement monitoring data well while ensuring a high recall rate, and reduce the false alarm rate of early warning.

Spike-Weighted Spiking Neural Network with Spiking Long Short-Term Memory: A Biomimetic Approach to Decoding Brain Signals

Background. Brain–machine interfaces (BMIs) offer users the ability to directly communicate with digital devices through neural signals decoded with machine learning (ML)-based algorithms. Spiking Neural Networks (SNNs) are a type of Artificial Neural Network (ANN) that operate on neural spikes instead of continuous scalar outputs. Compared to traditional ANNs, SNNs perform fewer computations, use less memory, and mimic biological neurons better. However, SNNs only retain information for short durations, limiting their ability to capture long-term dependencies in time-variant data. Here, we propose a novel spike-weighted SNN with spiking long short-term memory (swSNN-SLSTM) for a regression problem. Spike-weighting captures neuronal firing rate instead of membrane potential, and the SLSTM layer captures long-term dependencies. Methods. We compared the performance of various ML algorithms during decoding directional movements, using a dataset of microelectrode recordings from a macaque during a directional joystick task, and also an open-source dataset. We thus quantified how swSNN-SLSTM performed compared to existing ML models: an unscented Kalman filter, LSTM-based ANN, and membrane-based SNN techniques. Result. The proposed swSNN-SLSTM outperforms both the unscented Kalman filter, the LSTM-based ANN, and the membrane based SNN technique. This shows that incorporating SLSTM can better capture long-term dependencies within neural data. Also, our proposed swSNN-SLSTM algorithm shows promise in reducing power consumption and lowering heat dissipation in implanted BMIs.

A novel machine learning-based framework for the water quality parameters prediction using hybrid long short-term memory and locally weighted scatterplot smoothing methods

ABSTRACT Water quality prediction is crucial for effective river stream management. Dissolved oxygen, conductivity and chemical oxygen demand are vital chemical parameters for water quality. Development of machine learning (ML) and deep learning (DL) methods made them widely used in this domain. Sophisticated DL techniques, especially long short-term memory (LSTM) networks, are required for accurate, real-time multistep prediction. LSTM networks are effective in predicting water quality due to their ability to handle long-term dependencies in sequential data. We propose a novel hybrid approach for water quality parameters prediction combining DL with data smoothing method. The Sava river at the Jamena hydrological station serves as a case study. Our workflow uses LSTM networks alongside LOcally WEighted Scatterplot Smoothing (LOWESS) technique for data filtering. For comparison, Support Vector Regressor (SVR) is used as the baseline method. Performance is evaluated using Root Mean Squared Error (RMSE) and Coefficient of Determination R2 metrics. Results demonstrate that LSTM outperforms the baseline method, with an R2 up to 0.9998 and RMSE of 0.0230 on the test set for dissolved oxygen. Over a 5-day prediction period, our approach achieves R2 of 0.9912 and RMSE of 0.1610 confirming it as a reliable method for water quality multistep parameters prediction.

Ultra-short-term multi-energy load forecasting for integrated energy systems based on multi-dimensional coupling characteristic mining and multi-task learning

To address the challenges posed by the randomness and volatility of multi-energy loads in integrated energy systems for ultra-short-term accurate load forecasting, this paper proposes an ultra-short-term multi-energy load forecasting method based on multi-dimensional coupling feature mining and multi-task learning. Firstly, a method for mining multi-dimensional coupling characteristics of multi-energy loads is proposed, integrating multiple correlation analysis methods. By constructing coupling features of multi-energy loads and using them as input features of the model, the complex coupling relationships between multi-energy loads are effectively quantified. Secondly, an ultra-short-term multi-energy load forecasting model based on multi-task learning and a temporal convolutional network is constructed. In the prediction model construction phase, the potential complex coupling characteristics between multiple loads can be fully explored, and the potential temporal associations and long-term dependencies within data can be extracted. Then, the multi-task learning loss function weight optimization method based on homoscedastic uncertainty is used to optimize the forecasting model, realizing automatic tuning of the loss function weight parameters and further improving the prediction performance of the model. Finally, an experimental analysis is conducted using the integrated energy system of Arizona State University in the United States as an example. The results show that the proposed forecasting method has higher prediction accuracy than other prediction methods.

S2DA-Net: Spatial and spectral-learning double-branch aggregation network for liver tumor segmentation in CT images

Accurate liver tumor segmentation is crucial for aiding radiologists in hepatocellular carcinoma evaluation and surgical planning. While convolutional neural networks (CNNs) have been successful in medical image segmentation, they face challenges in capturing long-term dependencies among pixels. On the other hand, Transformer-based models demand a high number of parameters and involve significant computational costs. To address these issues, we propose the Spatial and Spectral-learning Double-branched Aggregation Network (S2DA-Net) for liver tumor segmentation. S2DA-Net consists of a double-branched encoder and a decoder with a Group Multi-Head Cross-Attention Aggregation (GMCA) module, Two branches in the encoder consist of a Fourier Spectral-learning Multi-scale Fusion (FSMF) branch and a Multi-axis Aggregation Hadamard Attention (MAHA) branch. The FSMF branch employs a Fourier-based network to learn amplitude and phase information, capturing richer features and detailed information without introducing an excessive number of parameters. The FSMF branch utilizes a Fourier-based network to capture amplitude and phase information, enriching features without introducing excessive parameters. The MAHA branch incorporates spatial information, enhancing discriminative features while minimizing computational costs. In the decoding path, a GMCA module extracts local information and establishes long-term dependencies, improving localization capabilities by amalgamating features from diverse branches. Experimental results on the public LiTS2017 liver tumor datasets show that the proposed segmentation model achieves significant improvements compared to the state-of-the-art methods, obtaining dice per case (DPC) 69.4 % and global dice (DG) 80.0 % for liver tumor segmentation on the LiTS2017 dataset. Meanwhile, the pre-trained model based on the LiTS2017 datasets obtain, DPC 73.4 % and an DG 82.2 % on the 3DIRCADb dataset.

Memory-Enhanced Knowledge Reasoning with Reinforcement Learning

In recent years, the emergence of large-scale language models, such as ChatGPT, has presented significant challenges to research on knowledge graphs and knowledge-based reasoning. As a result, the direction of research on knowledge reasoning has shifted. Two critical issues in knowledge reasoning research are the algorithm of the model itself and the selection of paths. Most studies utilize LSTM as the path encoder and memory module. However, when processing long sequence data, LSTM models may encounter the problem of long-term dependencies, where memory units of the model may decay gradually with an increase in time steps, leading to forgetting earlier input information. This can result in a decline in the performance of the LSTM model in long sequence data. Additionally, as the data volume and network depth increase, there is a risk of gradient disappearance. This study improved and optimized the LSTM model to effectively address the problems of gradient explosion and gradient disappearance. An attention layer was employed to alleviate the issue of long-term dependencies, and ConvR embedding was used to guide path selection and action pruning in the reinforcement learning inference model. The overall model achieved excellent reasoning results.

Fault detection of complicated processes based on an enhanced transformer network with graph attention mechanism

Recently, deep learning becomes increasingly popular in the industrial data analysis field due to its distinguished feature representation capability. As an emerging deep learning technology, transformer network has attracted extensive attention, but its application in industrial fault detection is still not sufficiently exploited. Furthermore, the traditional transformer network mainly focuses on the time series information of the data, while ignoring the spatial characteristic of the data. For this problem, an improved transformer network model with graph attention mechanism (GA-Tran) is proposed and used for complicated process fault detection. Considering the strong coupling property of process variables, the presented method performs joint spatial-temporal learning by mining the data information in views of both time-dimension and variable-dimension. On the one hand, a transformer encoder module equipped with a self-attention mechanism is utilized to capture long-term temporal dependencies, while multi-scale convolution is integrated to abstract short-term time dependencies. On the other hand, a graph attention network is introduced into the transformer network to meticulously analyze the spatial relationships among variables. Considering the topology of the industrial process variables, we use spectral clustering to infer prior topology information, and remove intelligently irrelevant spatial information. Different from the existing transformer models with the anomaly scores as the monitoring indicator, two monitoring statistics based on encoder features and reconstruction errors are constructed for process status monitoring. Studies on the Tennessee Eastman chemical process demonstrate the effectiveness of the proposed GA-Tran method.

Assessment of Quality of Life in Patients with Spinal Cord Injury within the Period of First 6 Months after Injury

Introduction: Spinal cord injury (SCI) is a devastating neurological state resulting in long-term physical dependency, increased morbidity, emotional, and financial burden. This study sought to assess the quality of life in patients with spinal cord injury within the first six months after injury. Aim of the study: The aim of this study was to assess the quality of life in patients with spinal cord injury within the first six months after injury. Methods: This cross-sectional study was conducted at the Department of Physical Medicine & Rehabilitation and Department of Neurosurgery in Bangabandhu Sheikh Mujib Medical University (BSMMU) over a period between October 2021 to September 2022. 21 patients with SCI within 6 months of injury were recruited. Interviews with a semi-structured questionnaire assessed demographics, SCI details, and QoL (WHOQOL-BREF). Neurological injury classification followed ASIA standards. Statistical analysis was performed using SPSS 24. Result: The majority of the studied patients (23.8%) were within the 41-50 and 51-60-year age groups. More than two-thirds of the studied patients (71.4%) were male. with falls (42.9%) and spinal tumors (19.0%) as leading causes. Among all, 33.3% of the patients had complete motor involvement. According to the score of the physical domain, 66.7% of patients had low QoL. Age was positively significantly correlated with the environmental QoL score (r: 0.532, p value: 0.013) Conclusion: SCI significantly impairs all aspects of quality of life, with complete SCI patients facing notably worse physical conditions. A collaborative approach involving health entities, families, and communities is essential to improve quality of life and self-care abilities for SCI survivors.

Optimizing Disease Prediction with Artificial Intelligence Driven Feature Selection and Attention Networks

The rapid integration of machine learning methodologies in healthcare has ignited innovative strategies for disease prediction, particularly with the vast repositories of Electronic Health Records (EHR) data. This article delves into the realm of multi-disease prediction, presenting a comprehensive study that introduces a pioneering ensemble feature selection model. This model, designed to optimize learning systems, combines statistical, deep, and optimally selected features through the innovative Stabilized Energy Valley Optimization with Enhanced Bounds (SEV-EB) algorithm. The objective is to achieve unparalleled accuracy and stability in predicting various disorders. This work proposes an advanced ensemble model that synergistically integrates statistical, deep, and optimally selected features. This combination aims to enhance the predictive power of the model by capturing diverse aspects of the health data. At the heart of the proposed model lies the SEV-EB algorithm, a novel approach to optimal feature selection. The algorithm introduces enhanced bounds and stabilization techniques, contributing to the robustness and accuracy of the overall prediction model. To further elevate the predictive capabilities, an HSC-AttentionNet is introduced. This network architecture combines deep temporal convolution capabilities with LSTM, allowing the model to capture both short-term patterns and long-term dependencies in health data. Rigorous evaluations showcase the remarkable performance of the proposed model. Achieving a 95% accuracy and 94% F1-score in predicting various disorders, the model surpasses traditional methods, signifying a significant advancement in disease prediction accuracy. The implications of this research extend beyond the confines of academia. By harnessing the wealth of information embedded in EHR data, the proposed model presents a paradigm shift in healthcare interventions. The optimized diagnosis and treatment pathways facilitated by this approach hold promise for more accurate and personalized healthcare, potentially revolutionizing patient outcomes

Evaluating the nonlinear association between PM10 and emergency department visits

Air pollution is one of the most threatening risk factors for human health. The epidemiological literature has widely proved that exposure to high levels of PM10 is associated with an increase in cardiovascular and respiratory events. In this paper, we investigate the relationship between air pollution and emergency department visits for cardiovascular and respiratory diseases. Since the relationship between air pollution and emergency care access is complex and may be subject to regime switches, especially due to the COVID-19 pandemic, we propose to use a threshold autoregressive model where the regime-switching mechanism is driven by temporal evolution. This allows us to identify one or more structural breaks and estimate the regime-changing relationship between emergency care accesses and PM10. Moreover, we propose a novel way to account for long-term dependence in the dependent variable. Using daily data from the metropolitan area of Bologna (Italy), we show that the effect of the air pollutant changes over time, mostly due to the outbreak of the COVID-19 pandemic.

Early fire detection technology based on improved transformers in aircraft cargo compartments

The implementation of early and accurate detection of aircraft cargo compartment fire is of great significance to ensure flight safety. The current airborne fire detection technology mostly relies on single-parameter smoke detection using infrared light. This often results in a high false alarm rate in complex air transportation environments. The traditional deep learning model struggles to effectively address the issue of long-term dependency in multivariate fire information. This paper proposes a multi-technology collaborative fire detection method based on an improved transformers model. Dual-wavelength optical sensors, flue gas analyzers, and other equipment are used to carry out multi-technology collaborative detection methods and characterize various feature dimensions of fire to improve detection accuracy. The improved Transformer model which integrates the self-attention mechanism and position encoding mechanism is applied to the problem of long-time series modeling of fire information from a global perspective, which effectively solves the problem of gradient disappearance and gradient explosion in traditional RNN (recurrent neural network) and CNN (convolutional neural network). Two different multi-head self-attention mechanisms are used to classify and model multivariate fire information, respectively, which solves the problem of confusing time series modeling and classification modeling in dealing with multivariate classification tasks by a single attention mechanism. Finally, the output results of the two models are fused through the gate mechanism. The research results show that, compared with the traditional single-feature detection technology, the multi-technology collaborative fire detection method can better capture fire information. Compared with the traditional deep learning model, the multivariate fire prediction model constructed by the improved Transformer can better detect fires, and the accuracy rate is 0.995.

Design and Implementation of Automatic Music Composition System Based on Deep Learning

Recently, there has been a notable increase in interest surrounding the utilization of Deep Learning Approaches for AMCS in recent times. However, current methods often lack coherence and authenticity in the generated music. This study proposes a new approach that utilizes RNNs to bridge the gap between traditional compositional methods and modern deep learning techniques. The goal is to produce expressive and coherent musical pieces. Our methodology involves designing and implementing a customized RNN architecture that can effectively capture the complex temporal dependencies present in musical sequences. We experiment with different types of including LSTM networks, RNNs, and GRUs, to overcome challenges such as vanishing gradients and better model longer-term dependencies in music. To train the neural network efficiently and improve model convergence, various deep learning optimizers are utilized in our system. Specifically, we use SGD optimizer to improve the hyper parameters of LSTM and GRUs. The data for training is converted into MIDI format and analyzed, with music lines being identified through a similarity matrix technique. The MIDI data is then prepared for use in the LSTM and GRUs networks. The resulting music is assessed using both objective measures, such as mean squared error, and subjective approaches. This research adds to the advancements in automatic music composition by demonstrating the capability of RNNs in capturing and producing complex musical arrangements.

Oncological and functional outcome after laryngectomy for laryngeal and hypopharyngeal cancer: a population-based analysis in Germany from 2001 to 2020

Prognostic factors for overall survival (OS), percutaneous endoscopic gastrostomy (PEG) dependency, and long-term speech rehabilitation via voice prosthesis (VP) after laryngectomy for laryngeal or hypopharyngeal cancer were investigated in a retrospective population-based study in Thuringia, Germany. A total of 617 patients (68.7% larynx; hypopharynx; 31.3%; 93.7% men; median age 62 years; 66.0% stage IV) from 2001 to 2020 were included. Kaplan–Meier and Cox multivariable regression analyses were performed. 23.7% of patients received a PEG. 74.7% received a VP. Median OS was 131 months. Independent factors for lower OS were stage IV (compared to stage II; hazard ratio [HR] = 3.455; confidence interval [CI] 1.395–8.556) and laryngectomy for a recurrent disease (HR = 1.550; CI 1.078–2.228). Median time to PEG removal was 7 months. Prior partial surgery before laryngectomy showed a tendency for independent association for later PEG removal (HR = 1.959; CI 0.921–4.167). Postoperative aspiration needing treatment was an independent risk factor (HR = 2.679; CI 1.001–7.167) for later definitive VP removal. Laryngectomy continuously plays an important role in a curative daily routine treatment setting of advanced laryngeal or hypopharyngeal cancer in Germany. Long-term dependency on nutrition via PEG is an important issue, whereas use of VP is a stable long-term measure for voice rehabilitation.

A Novel Hybrid Model Combining BPNN Neural Network and Ensemble Empirical Mode Decomposition

Neural network models have been successfully used to predict stock prices, weather, and traffic patterns. Due to the sensitivity of the data, it is very effective in identifying and maintaining long-term dependencies in time series. The back propagation neural network (BPNN) model works well in regression and classification applications, such as predicting stock prices and sales volumes. BPNN needs to sort out the mapping between inputs and outputs before continuous values. BPNN neural network model is integrated with ensemble empirical mode decomposition (EEMD), and a new hybrid neural network prediction model is constructed. Integrating ensemble empirical mode decomposition, collecting and preprocessing sequence features, reducing noise, improving robustness, and then training neural networks with returned feature vectors instead. In the international gold price series forecasting, the R2\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R^2$$\\end{document} of the new hybrid model is 1.85%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document} better than the existing EEMD-LSTM model, 3.8%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document} and 5.44%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document} better than the independent BPNN and long short-term memory network (LSTM) neural network models, respectively. Compared with LSTM, the BPNN plays the performance of EEMD better, reduces the error to a certain extent, and improves the prediction accuracy.

Quantitative blood glucose detection influenced by various factors based on the fusion of photoacoustic temporal spectroscopy with deep convolutional neural networks

In order to efficiently and accurately monitor blood glucose concentration (BGC) synthetically influenced by various factors, quantitative blood glucose in vitro detection was studied using photoacoustic temporal spectroscopy (PTS) combined with a fusion deep neural network (fDNN). Meanwhile, a photoacoustic detection system influenced by five factors was set up, and 625 time-resolved photoacoustic signals of rabbit blood were collected under different influencing factors.In view of the sequence property for temporal signals, a dimension convolutional neural network (1DCNN) was established to extract features containing BGC. Through the parameters optimization and adjusting, the mean square error (MSE) of BGC was 0.51001 mmol/L for 125 testing sets. Then, due to the long-term dependence on temporal signals, a long short-term memory (LSTM) module was connected to enhance the prediction accuracy of BGC. With the optimal LSTM layers, the MSE of BGC decreased to 0.32104 mmol/L. To further improve prediction accuracy, a self-attention mechanism (SAM) module was coupled into and formed an fDNN model, i.e., 1DCNN-SAM-LSTM. The fDNN model not only combines the advantages of temporal expansion of 1DCNN and data long-term memory of LSTM, but also focuses on the learning of more important features of BGC. Comparison results show that the fDNN model outperforms the other six models. The determination coefficient of BGC for the testing set was 0.990, and the MSE reached 0.1432 mmol/L. Results demonstrate that PTS combined with 1DCNN-SAM-LSTM ensures higher accuracy of BGC under the synthetical influence of various factors, as well as greatly enhances the detection efficiency.