Prediction of pollution loads in the Geum River upstream using the recurrent neural network algorithm

With the rapid development of neural network technology, we have widely used this technology in various fields. In the field of language translation, the research on automatic detection technology of English verb grammatical errors is in a hot stage. The traditional manual detection cannot be applied to the current environment. Therefore, this paper proposes an automatic detection technology of English verb grammatical errors based on recurrent neural network (RNN) algorithm to solve this problem. Firstly, the accuracy and feedback speed of traditional manual detection and recurrent neural network RNN algorithm are compared. Secondly, a detection model which can be calculated according to grammatical order combined with context is designed. Finally, when the output verb result is inconsistent with the original text, it can automatically mark the error detection effect. The experimental results show that the algorithm model studied in this paper can effectively improve the detection accuracy and feedback efficiency and is more applicable and effective than the traditional manual detection method.

A real-coded genetic algorithm for training recurrent neural networks

In an effort to improve energy efficiency and sustainability in the agricultural sector, smart technology has been integrated into the greenhouse system. The research utilizes the Recurrent Neural Network (RNN) algorithm to forecast values of irradiance on a time principal. The RNN algorithm is chosen for its ability to handle time-series data and predict patterns based on historical data. By using the RNN algorithm, the system can predict real-time needs and then use this information to optimally distribute power from solar power plants. Additionally, this system is equipped with Internet of Things (IoT)-based monitoring capabilities, allowing remote monitoring and control of the research object. Connected IoT sensors collect real-time environmental data and send it to the data server for analysis. The data is also used to update the model of RNN, making supply prediction more accurate over time. The implementation results show increased energy efficiency and reduced operational costs in Green House management. By leveraging AI and IoT technology, model evaluation is conducted using RMSE, MSE, MAE, and R-squared (R²) metrics as important indicators of model accuracy. The evaluation results indicate that this model can provide accurate predictions of irradiance patterns, with low RMSE and MAE values and R² approaching one, signifying excellent implementation in capturing data dynamics.

This article proposes an English grammar intelligent error correction model based on the attention mechanism and Recurrent Neural Network (RNN) algorithm. It aims to improve the accuracy and effectiveness of error correction by combining the powerful context-capturing ability of the attention mechanism with the sequential modeling ability of RNN. First, based on the improvement of recurrent neural networks, a bidirectional gated recurrent network is added to form a dual encoder structure. The encoder is responsible for reading and understanding the input text, while the decoder is responsible for generating the corrected text. Second, the attention mechanism is introduced into the decoder to convert the output of the encoder into the attention probability distribution for integration. This allows the model to focus on the relevant input word as it generates each corrected word. The results of the study showed that the model was 2.35% points higher than statistical machine translation–neural machine translation in the CoNLL-2014 test set, and only 1.24 points lower than the human assessment score, almost close to the human assessment level. The model proposed in this study not only created a new way of English grammar error correction based on the attention mechanism and RNN algorithm in theory but also effectively improved the accuracy and efficiency of English grammar error correction in practice. It further provides English learners with higher-quality intelligent error correction tools, which can help them learn and improve their English level more effectively.

This paper presents a novel data-based fault diagnosis approach of aircraft actuators by using deep learning methods. Electro-mechanical actuator (EMA), which we study on, widely used in a new generation of aircraft serves as the research object. The basic fault diagnosis framework of this work is based on Time Series Modeling (TSM), which builds a model for each sensor data, then faults can be detected and isolated by differences between model prediction result and measured result. In this structure, the quantity of sensors is directly related to the fault diagnosis effect, that is, the fault detection capability increases when the number of sensors increases. Simultaneously, the workload of the model training increases without expectation. Therefore, the reduction of sensors is of great significance. The traditional method is to use the manual experience to screen, but this method has high requirements on personal ability and performance is not easy to guarantee. This paper adopts the Sparse Auto-Encoder (SAE) algorithm, which avoids the simple direct screening process and achieves the purpose of reducing the number of sensors by adaptively extracting features. Important features of sensor data, preserved by loss compression using SAE, can be used to build time series models. Relationship of time series adjacent data is adopted to build mathematical models. Compared with traditional machine learning algorithms, recurrent neural network (RNN) algorithm can make good use of the relationship between time series, which has been widely used in speech recognition, text recognition, and other fields. However, standard RNN algorithm tends to ignore future information. In this paper, the Bidirectional Long Short-term Memory RNN algorithm (BiLSTM-RNN), considering both past data and future data, is applied to TSM. Through the Comparison of the standard RNN and long short-term memory RNN (LSTM-RNN) algorithms, the biLSTM-RNN algorithm shows better modeling and fault diagnosis performance.

Artificial Intelligence of Things (AIoT) is a new computing technology that combines Artificial Intelligence (AI) and Internet of Things (AIoT) technologies that are used to improve data analysis systems on implanted devices. IoT can be an innovative solution in the development of more efficient embedded devices This research will delineate the description of the AIoT system using the Recurrent Neural Network (RNN) algorithm. Weather parameters used as input are humidity, temperature, wind direction, wind speed, and atmospheric pressure. Based on the insert received from the sensor, the accuracy of the RNN algorithm, will be known to the prediction results of weather conditions in oil palm plantations. The purpose of this research is to develop an IoT device with an artificial neural network algorithm embedded in the device to predict weather conditions in real-time.

The objective of this work is to compare the Recurrent Neural Network (RNN) algorithm and Support Vector Machine (SVM) algorithm in the identification of endometrial cancer based on its accuracy and sensitivity measurements. Materials and Methods: The endometrial cancer dataset, obtained from the National Institute of Endometrial Cancer Diseases (NIECE), contains 768 patient health records that were used to train (80 %) and test (20 %) the predictive model in MATLAB and the statistical analysis is done using SPSS software. For this research work 768 images were used with the pixel size of 3048×2048 and these images are taken from the pap smear slide dataset. The RNN algorithm is used and compared with the SVM algorithm. The sample size is estimated for two groups (RNN & SVM) with G-power of 80 % and 0.05 Type I/II Error rate (Alpha). Results: The predictive model using RNN algorithm shows a higher accuracy of 93.90 ± 0.3160 and sensitivity of 91.0400 ± 1.07207 followed by the significance value of 0.002 than SVM algorithm with accuracy of 88.10 ± 0.9940 and sensitivity of 86.1700 ± 1.36793 with the significance value of 0.000 using 2-tailed test in SPSS. Conclusion: Based on the outcome of the proposed work RNN classifier shows significantly better performance than the SVM classifier in the innovative detection of endometrial cancer.

This thesis deals with recurrent neural networks, a particular class of artificial neural networks which can learn a generative model of input sequences. The input is mapped, through a feedback loop and a non-linear activation function, into a hidden state, which is then projected into the output space, obtaining either a probability distribution or the new input for the next time-step. This work consists mainly of two parts: a theoretical study for helping the understanding of recurrent neural networks framework, which is not yet deeply investigated, and their application to non-linear prediction problems, since recurrent neural networks are really powerful models suitable for solving several practical tasks in different fields. For what concerns the theoretical part, we analyse the weaknesses of state-of-the-art models and tackle them in order to improve the performance of a recurrent neural network. Firstly, we contribute in the understanding of the dynamical properties of a recurrent neural network, highlighting the close relation between the definition of stable limit cycles and the echo state property of an echo state network. We provide sufficient conditions for the convergence of the hidden state to a trajectory, which is uniquely determined by the input signal, independently of the initial states. This may help extend the memory of the network and increase the design options for the network. Moreover, we develop a novel approach to address the main problem in training recurrent neural networks, the so-called vanishing gradient problem. Our new method allows us to train a very simple recurrent neural network, making the gradient not to vanish even after many time-steps. Exploiting the singular value decomposition of the vanishing factors in the gradient and random matrices theory, we find that the singular values have to be confined in a narrow interval and derive conditions about their root mean square value. Then, we also improve the efficiency of the training of a recurrent neural network, defining a new method for speeding up this process. Thanks to a least square regularization, we can initialize the parameters of the network, in order to set them closer to the minimum and running fewer epochs of classical training algorithms. Moreover, it is also possible to completely train the network with our initialization method, running more iterations of it without losing in performance with respect to classical training algorithms. Finally, it is also possible to use it as a real-time learning algorithm, adjusting the parameters to the new data through one iteration of our initialization. In the last part of this thesis, we apply recurrent neural networks to non-linear prediction problems. We consider prediction of numerical sequences, estimating the following input choosing it from a probability distribution. We study an automatic text generation problem, where we need to predict the following character in order to compose words and sentences, and a path prediction of walking mobile users in the central area of a city, as a sequence of crossroads. Then, we analyse the prediction of video frames, discovering a wide range of applications related to the prediction of movements. We study the collision problem of bouncing balls, taking into account only the sequence of video frames without any knowledge about the physical characteristics of the problem, and the distribution over days of mobile user in a city and in a whole region. Finally, we address the state-of-the-art problem of missing data imputation, analysing the incomplete spectrogram of audio signals. We restore audio signals with missing time-frequency data, demonstrating via numerical experiments that a performance improvement can be achieved involving recurrent neural networks.

The purpose of this study is to evaluate the performance of a recurrent neural network (RNN)-based prediction algorithm to compensate for respiratory movement using an articulated robotic couch system. A prototype of a real-time respiratory motion compensation couch was built using an optical 3D motion tracking system and a six-degree-of-freedom-articulated robotic system. To compensate for the system latency from motion detection to re-positioning of the system, RNN and double exponential smoothing (ES2) prediction algorithms were applied. Three aspects of performance were evaluated, simulation and experiments for geometric and dosimetric evaluations, using data from three liver and three lung patients who underwent stereotactic body radiotherapy. Overall, the RNN algorithm showed better geometric and dosimetric results than the other approaches. In simulation tests, RNN showed 82% average improvement ratio, compared with non-predicted results. In the geometric evaluation, RNN only showed average FWHM broadening of 1.5 mm, compared with the static case. In the dosimetric evaluation, RNN showed average gamma passing rates of 97.4 ± 1.0%, 89.0 ± 2.4% under the 3%/3 mm, 2%/2 mm respectively. It may be technically feasible to use the RNN prediction algorithm to compensate for respiratory motion with an articulated robotic couch system. The RNN algorithm could be widely used for motion compensation in patients undergoing radiotherapy.

Cultivated land quality (CLQ) is associated with national food security, benign economic development, social harmony, and stability. The scientific evaluation of CLQ provides the basis for achieving the “trinity” protection of cultivated land quantity, and quality, as well as ecology. However, the current research on CLQ evaluation has some limitations, mainly the poor consideration of evaluation indicators, time-consuming and labor-intensive data acquisition, and low precision of evaluation at the regional scale. Therefore, this study introduced multisource data to evaluate CLQ and proposed a new method for CLQ evaluation (natural grade evaluation, utilization grade evaluation, and economic grade evaluation), combining multisource data and the recurrent neural network (RNN) algorithm. Initially, optimal indicators were determined by correlation analysis and generalized linear regression coefficient methods based on factors related to CLQ acquired from multisource data. Then, CLQ evaluation models were constructed with the RNN algorithm on the basis of the aforementioned optimal indicators. Finally, the models were adopted to map CLQ. The present study was carried out in Guangzhou City, Guangdong Province, China. According to the results: (1) CLQ showed close relationship to pH, effective soil layer thickness (EST), chemical fertilizer application rate (CHFE), organic matter content (OMC), annual accumulated temperature (TEMA), 5–15 cm soil depth soil cation exchange capacity (CEC515), 0–5 cm soil depth soil cation exchange capacity (CEC05), 5–15 cm soil depth soil organic carbon content (SOC515), 0–5 cm soil depth soil organic carbon content (SOC05), field slope (FS), groundwater level (GWL), and terrain slope (TS). (2) All modeling accuracies (R2) were greater than 0.80 for the CLQ evaluation models constructed based on the RNN algorithm. The area and spatial distribution of each grade of CLQ evaluation were consistent with the actual situation. The best and the worst quality cultivated land occupied a small area, and the area without a gap with the actual CLQ was as high as 76%, indicating that the model results were reliable. The study shows the suitability of the method for evaluating CLQ at the regional scale, offering a scientific foundation for the rational utilization and management of cultivated land resources, as well as a reference for evaluating CLQ in the future.

Evaluation of deep learning algorithms for national scale landslide susceptibility mapping of Iran

A generalized LSTM-like training algorithm for second-order recurrent neural networks.

This paper aims to develop an accurate estimation technique for computing state of charge (SOC) of a lithium-ion battery using recurrent neural network algorithm. Nonlinear autoregressive with exogenous input (NARX) model is a well-known subclass of the recurrent neural network which has proven to be very effective and computationally rich for controlling dynamic system and predicting time series. However, the accuracy of recurrent NARX neural network depends on the amount of input and output order as well as a number of neurons in a hidden layer. Therefore, this study presents an improved recurrent NARX neural network based SOC estimation with particle swarm optimization (PSO) algorithm for finding the best value of input delays, feedback delays and a number of neurons in a hidden layer. The proposed model uses three most significant factor such as current, voltage and temperature without considering battery model. The model robustness is checked at low temperature (0°C), medium temperature (25°C), and high temperature (45°C). The US06 drive cycle is selected for model training and testing. The effectiveness of the proposed approach is compared with the back-propagation neural network (BPNN) optimized by PSO based on the SOC error, root mean square error (RMSE) and mean absolute error (MAE) and average execution time (AET). The results prove that the proposed model has higher estimation speed and achieves higher accuracy in reducing RMSE and MAE by 53% and 50% than BPNN based PSO model at 25°C.

Aim: The main aim of the study is to predict metro water fraud accurately by Recurrent Neural Network Algorithms and compare the prediction accuracy with Convolutional Neural Network. Materials and Methods: In the existing system Convolutional Neural Network algorithm is used and in the proposed system Recurrent Neural Network algorithm is used. CNN with sample size =20 and RNN with sample size =20 was iterated forty times for predicting the accuracy. The algorithms have been implemented and tested over a dataset which consists of 8002 records. Result: After performing the experiment we get mean accuracy of 94.5210 by using Recurrent Neural Network algorithm and we get accuracy of 93.4950 by using Convolutional Neural Network algorithm for metro water fraudulent prediction. There is a statistical significant difference in accuracy for two algorithms is p<0.05 by performing independent samples t-tests. Conclusion: The comparison results show that the Recurrent Neural Network algorithm appears to be better performance than Convolutional Neural Network algorithms.

To achieve imaging of objects using cosmic ray muons, it is essential to determine the muon trajectories, and accurately locating the muon hit points is key to trajectory reconstruction. Current mainstream muon track detection systems require multiple electronic channels to precisely pinpoint the muon hit points, resulting in complex structures and high costs. The aim of this study is to design a muon track detection system characterized by simplicity, low cost, and high precision. This system is based on Geant4 software and features a circular plastic scintillator without segmentation coupled with Silicon Photon Multipliers (SiPMs). The simulation involves the use of SiPMs to collect the number of photons and the time of triggering SiPM responses as characteristic parameters, and employs a recurrent neural network (RNN) algorithm as the method for muon localization. The simulation results show that a plastic scintillator with a diameter of 20 cm and a thickness of 10 mm can achieve a position resolution of less than 1 cm while ensuring a large detection area. The more SiPMs are coupled, the more accurate the collected photon information is, which improves the position resolution of the detector. The reconstruction accuracy of the Gated Recurrent Unit (GRU) algorithm is comparable to that of the Long Short-Term Memory (LSTM) algorithm within the uncertainties, but the reconstruction speed of the GRU algorithm is better than that of the LSTM algorithm. Under the same conditions, the training speed of the GRU algorithm is about 9/11 of that of the LSTM algorithm. The results of the multi-layer detection unit optimization scheme proposed in this study indicate that as the minimum number of detection unit layers in the muon positioning module increases, the position resolution of the system gradually improves. Considering the cost, the GRU algorithm can be chosen as the reconstruction algorithm, and the minimum detection unit with 8 SiPMs coupled in two layers can be selected as the system optimization scheme. Considering the accuracy of position resolution, the LSTM algorithm can be chosen as the reconstruction algorithm, and the minimum detection unit with 6 SiPMs coupled in three layers can be selected as the system optimization scheme. The multi-unit integrated muon detection system proposed in this study ensures a large detection area while enabling the simplest minimum detection units in a multilayer structure to work in coordination, thereby optimizing the position resolution of the entire muon detection system.