Number Of Hidden Layers Research Articles

Credit Card Fraud Detection Using State-of-the-Art Machine Learning and Deep Learning Algorithms

People can use credit cards for online transactions as it provides an efficient and easy-to-use facility. With the increase in usage of credit cards, the capacity of credit card misuse has also enhanced. Credit card frauds cause significant financial losses for both credit card holders and financial companies. In this research study, the main aim is to detect such frauds, including the accessibility of public data, high-class imbalance data, the changes in fraud nature, and high rates of false alarm. The relevant literature presents many machines learning based approaches for credit card detection, such as Extreme Learning Method, Decision Tree, Random Forest, Support Vector Machine, Logistic Regression and XG Boost. However, due to low accuracy, there is still a need to apply state of the art deep learning algorithms to reduce fraud losses. The main focus has been to apply the recent development of deep learning algorithms for this purpose. Comparative analysis of both machine learning and deep learning algorithms was performed to find efficient outcomes. The detailed empirical analysis is carried out using the European card benchmark dataset for fraud detection. A machine learning algorithm was first applied to the dataset, which improved the accuracy of detection of the frauds to some extent. Later, three architectures based on a convolutional neural network are applied to improve fraud detection performance. Further addition of layers further increased the accuracy of detection. A comprehensive empirical analysis has been carried out by applying variations in the number of hidden layers, epochs and applying the latest models. The evaluation of research work shows the improved results achieved, such as accuracy, f1-score, precision and AUC Curves having optimized values of 99.9%,85.71%,93%, and 98%, respectively. The proposed model outperforms the state-of-the-art machine learning and deep learning algorithms for credit card detection problems. In addition, we have performed experiments by balancing the data and applying deep learning algorithms to minimize the false negative rate. The proposed approaches can be implemented effectively for the real-world detection of credit card fraud

Depth-based vessel position fixing by means of a neural network

A depth-based vessel position fixing method on the basis of a neural network is proposed. The network takes as an input a sequence of depth values measured by an echo-sounder and predicts vessel latitude and longitude for the moment of the latest depth measurement. The neural network has a fully-connected feedforward architecture with several layers which satisfies conditions of the universal approximation in compliance with the Stone-Weierstrass theorem. The Adamax algorithm for the neural network training with controlling a maximum value of position error at each epoch is implemented. Modeling is conducted with the Python programming language and the Tensorflow library. The model surface of seabed is performed as a second-order polynomial. Training samples on the basis of virtual soundings at the coordinate net knots with the space resolution not worse than one cable are obtained. After samples obtaining the training of the neural network is conducted. A validation set is not used. Several neural networks are trained. They have different number of hidden layers and different number of neurons per each hidden layer. After training the test procedure is performed. Test samples are generated in the assumption that a vessel is moving along meridians which are not used at the stage of the preliminary soundings survey. The cases of mean and random test meridians are considered. The random meridians are obtained with a uniform random number generator. As the result, all the tested neural networks have shown approximately identical navigational accuracy which is close to the accuracy for the training set.

Artificial intelligence-based prediction of the rheological properties of hydrocolloids for plant-based meat analogues.

Methylcellulose has been applied as a primary binding agent to control the quality attributes of plant-based meat analogues. H owever, a great deal of effort has been made to search for hydrocolloids to replace methylcellulose because of increasing awareness of clean labels. In this study, a machine learning framework was proposed in order to describe and predict the flow behavior of six hydrocolloid solutions, and the predicted viscosities were correlated with the textural features of their corresponding plant-based meat analogues. Different shear-thinning and Newtonian behaviors were observed depending on the type of hydrocolloid and the shear rate. Methylcellulose exhibited an increasing viscosity pattern with increasing temperature, compared to the other hydrocolloids. The machine learning algorithms (random forest and multilayer perceptron models) showed a better viscosity fitting performance than the constitutive equations (power law and Cross models). In addition, three hyperparameters of the multilayer perceptron model (optimizer, learning rate, and the number of hidden layers) were tuned using the Bayesian optimization algorithm. The optimized multilayer perceptron model exhibited superior performance in viscosity prediction (R2 = 0.9944-0.9961/RMSE = 0.0545-0.0708). Furthermore, the machine learning-predicted viscosities overall showed similar patterns to the textural parameters of the meat analogues. © 2024 Society of Chemical Industry.

Enhancing the Design of Experiments on the Fatigue Life Characterisation of Fibre-Reinforced Plastics by Incorporating Artificial Neural Networks.

Fatigue life testing is a complex and costly matter, especially in the case of fibre-reinforced thermoplastics, where other parameters in addition to force alone must be taken into account. The number of tests required therefore increases significantly, especially if the influence of different fibre orientations is to be taken into account. It is therefore important to gain the greatest possible amount of knowledge from the limited number of available tests. In order to achieve this, this study aims to utilise adaptive sampling, which is used in numerous areas of computational engineering, for the design of experiments on fatigue life testing. Artificial neural networks (ANNs) are therefore trained on data for the short-fibre-reinforced material PBT GF30, and their areas of greatest model uncertainty are queried. This was undertaken with ANNs from various numbers of hidden layers, which were analysed for their performance. The ideal case turned out to be four hidden layers, for which a squared error as small as 1 × 10-3 was recorded. Locally resolved, the ANN was used to identify the region of greatest uncertainty for samples of vertical orientation and small numbers of cycles. With information such as this, additional data can be obtained in such uncertain regions in order to improve the model prediction-almost halving the recorded error to only 0.55 × 10-3. In this way, a model of comparable value can be found with less experimental effort, or a model of better quality can be set up with the same experimental effort.

Impact of large kernel size on yield prediction: a case study of corn yield prediction with SEDLA in the U.S. Corn Belt

Predicting agricultural yields is imperative for effective planning to sustain the growing global population. Traditionally, regression-based, simulation-based, and hybrid methods were employed for yield prediction. In recent times, there has been a notable shift towards the adoption of Machine Learning (ML) methods, with Deep Learning (DL), particularly Convolutional Neural Networks (CNNs) and Long-Short Term Memory (LSTM) networks, emerging as popular choices for their enhanced predictive accuracy. This research introduces a cost-effective DL architecture tailored for corn yield prediction, considering computational efficiency in processing time, data size, and NN architecture complexity. The proposed architecture, named SEDLA (Simple and Efficient Deep Learning Architecture), leverages the spatial and temporal learning capabilities of CNNs and LSTMs, respectively, with a unique emphasis on exploring the impact of kernel size in CNNs. Simultaneously, the study aims to exclusively employ satellite and yield data, strategically minimizing input variables to enhance the model’s simplicity and efficiency. Notably, the study demonstrates that employing larger kernel sizes in CNNs, especially when processing histogram-based Surface Reflectance (SR) and Land Surface Temperature (LST) data from Moderate Resolution Imaging Spectroradiometer (MODIS), allows for a reduction in the number of hidden layers. The efficacy of the architecture was evaluated through extensive testing on corn yield prediction across 13 states in the United States (U.S.) Corn Belt at county-level. The experimental results showcase the superiority of the proposed architecture, achieving a Mean Absolute Percentage Error (MAPE) of 6.71 and Root Mean Square Error (RMSE) of 14.34, utilizing a single-layer CNN with a 15 × 15 kernel in conjunction with LSTM. These outcomes surpass existing benchmarks in the literature, affirming the efficacy and potential of the suggested DL framework for accurate and efficient crop yield predictions.

Voiceprint Fault Diagnosis of Converter Transformer under Load Influence Based on Multi-Strategy Improved Mel-Frequency Spectrum Coefficient and Temporal Convolutional Network

In order to address the challenges of low recognition accuracy and the difficulty in effective diagnosis in traditional converter transformer voiceprint fault diagnosis, a novel method is proposed in this article. This approach takes account of the impact of load factors, utilizes a multi-strategy improved Mel-Frequency Spectrum Coefficient (MFCC) for voiceprint signal feature extraction, and combines it with a temporal convolutional network for fault diagnosis. Firstly, it improves the hunter–prey optimizer (HPO) as a parameter optimization algorithm and adopts IHPO combined with variational mode decomposition (VMD) to achieve denoising of voiceprint signals. Secondly, the preprocessed voiceprint signal is combined with Mel filters through the Stockwell transform. To adapt to the stationary characteristics of the voiceprint signal, the processed features undergo further mid-temporal processing, ultimately resulting in the implementation of a multi-strategy improved MFCC for voiceprint signal feature extraction. Simultaneously, load signal segmentation is introduced for the diagnostic intervals, forming a joint feature vector. Finally, by using the Mish activation function to improve the temporal convolutional network, the IHPO-ITCN is proposed to adaptively optimize the size of convolutional kernels and the number of hidden layers and construct a transformer fault diagnosis model. By constructing multiple sets of comparison tests through specific examples and comparing them with the traditional voiceprint diagnostic model, our results show that the model proposed in this paper has a fault recognition accuracy as high as 99%. The recognition accuracy was significantly improved and the training speed also shows superior performance, which can be effectively used in the field of multiple fault diagnosis of converter transformers.

Evaluation method for psychological resilience of athletes in high-intensity sports training based on evolutionary neural network

To enhance the psychological resilience of athletes, a method for evaluating the psychological resilience of High-intensity Interval Training (HIIT) athletes based on evolutionary neural networks is studied. From the six criteria of frustration coping, personal characteristics, self-promotion, self-regulation, internal protection and external protection, the evaluation index of psychological resilience of athletes in sports High-intensity Interval Training is selected; the audition indicators are qualitatively analyzed according to the principle of indicator selection, and the indicators that do not meet the requirements are eliminated; Cluster analysis and coefficient of variation analysis are used to carry out quantitative analysis on the remaining evaluation indicators after qualitative analysis; the indicators after quantitative analysis are improved, to build the assessment index system of psychological resilience of athletes in high-intensity sports training. The Back Propagation (BP) neural network is optimized by a genetic algorithm, and the evolutionary neural network is constructed. The index data set is input into the evolutionary neural network as a sample, and the index weight value is output through training. The evaluation result and corresponding evaluation grade are determined based on the index weight value and membership degree. The experimental results show that when the number of hidden layers is 3, the calculation of evaluation index weights is the best; The weight of personal traits obtained from the evaluation results is the highest (0.206), while the weight of external protection is the lowest (0.151), and the evaluation results are basically consistent with the expert results. The above results show that this method can accurately evaluate the psychological resilience of athletes and significantly enhance their psychological resilience.

Machine learning classification of cellular states based on the impedance features derived from microfluidic single-cell impedance flow cytometry.

Mitosis is a crucial biological process where a parental cell undergoes precisely controlled functional phases and divides into two daughter cells. Some drugs can inhibit cell mitosis, for instance, the anti-cancer drugs interacting with the tumor cell proliferation and leading to mitosis arrest at a specific phase or cell death eventually. Combining machine learning with microfluidic impedance flow cytometry (IFC) offers a concise way for label-free and high-throughput classification of drug-treated cells at single-cell level. IFC-based single-cell analysis generates a large amount of data related to the cell electrophysiology parameters, and machine learning helps establish correlations between these data and specific cell states. This work demonstrates the application of machine learning for cell state classification, including the binary differentiations between the G1/S and apoptosis states and between the G2/M and apoptosis states, as well as the classification of three subpopulations comprising a subgroup insensitive to the drug beyond the two drug-induced states of G2/M arrest and apoptosis. The impedance amplitudes and phases used as input features for the model training were extracted from the IFC-measured datasets for the drug-treated tumor cells. The deep neural network (DNN) model was exploited here with the structure (e.g., hidden layer number and neuron number in each layer) optimized for each given cell type and drug. For the H1650 cells, we obtained an accuracy of 78.51% for classification between the G1/S and apoptosis states and 82.55% for the G2/M and apoptosis states. For HeLa cells, we achieved a high accuracy of 96.94% for classification between the G2/M and apoptosis states, both of which were induced by taxol treatment. Even higher accuracy approaching 100% was achieved for the vinblastine-treated HeLa cells for the differentiation between the viable and non-viable states, and between the G2/M and apoptosis states. We also demonstrate the capability of the DNN model for high-accuracy classification of the three subpopulations in a complete cell sample treated by taxol or vinblastine.

Color-Based Point Cloud Classification Using a Novel Gaussian Mixed Modeling-Based Approach versus a Deep Neural Network

The classification of point clouds is an important research topic due to the increasing speed, accuracy, and detail of their acquisition. Classification using only color is basically absent in the literature; the few available papers provide only algorithms with limited usefulness (transformation of three-dimensional color information to a one-dimensional one, such as intensity or vegetation indices). Here, we proposed two methods for classifying point clouds in RGB space (without using spatial information) and evaluated the classification success since it allows a computationally undemanding classification potentially applicable to a wide range of scenes. The first is based on Gaussian mixture modeling, modified to exploit specific properties of the RGB space (a finite number of integer combinations, with these combinations repeated in the same class) to automatically determine the number of spatial normal distributions needed to describe a class (mGMM). The other method is based on a deep neural network (DNN), for which different configurations (number of hidden layers and number of neurons in the layers) and different numbers of training subsets were tested. Real measured data from three sites with different numbers of classified classes and different “complexity” of classification in terms of color distinctiveness were used for testing. Classification success rates averaged 99.0% (accuracy) and 96.2% (balanced accuracy) for the mGMM method and averaged 97.3% and 96.7% (balanced accuracy) for the DNN method in terms of the best parameter combinations identified.

Deep learning to develop zero-equation based turbulence model for CFD simulations of the built environment

This study aims to improve the accuracy and speed of predictions for thermal comfort and air quality in built environments by creating a coupled framework between computational fluid dynamics (CFD) simulations and deep learning models. The coupling approach is showcased by the development of a data-driven turbulence model. The new turbulence model is built using a deep learning neural network, whose mapping structure is based on a zero-equation turbulence model for built environment simulations, and is coupled with the CFD software OpenFOAM to create a hybrid framework. The neural network is a standard shallow multi-layer perceptron. The number of hidden layers and nodes per layer was optimized using Bayesan optimization algorithm. The framework is trained on an indoor environment case study, as well as tested on an indoor office simulation and an outdoor building array simulation. Results show that the deep learning based turbulence model is more robust and faster than traditional two-equation Reynolds average Navier-Stokes (RANS) turbulence models, while maintaining a similar level of accuracy. The model also outperforms the standard algebraic zero-equation model due to its superior ability to generalize to various flow scenarios. Despite some challenges, namely the mapping constraint, the limited training dataset size and the source of generation of training data, the hybrid framework demonstrates the viability of the coupling technique and serves as a starting point for future development of more reliable and advanced models.

Do we need a more sophisticated multilayer artificial neural network to compute roughness coefficient?

Artificial neural networks (ANNs) are one of the most rapidly growing fields of soft computing. Along with deep learning, they are currently the most widely used machine learning techniques. Artificial neural networks are especially suitable for problem-solving where a researcher deals with incomplete data sets and no algorithms or specific sets of rules to be followed.This article deals with a case of comparison of several modifications of neural networks that may be applied to compute Chézy’s roughness coefficient. Neural network modelling is often started with one hidden layer. Having even one hidden layer, a neural network presents a powerful computing system to give good results. If it is necessary, the number of hidden layers may increase. Usually, two or three hidden layers of neurons are used. Diverse activation functions may also apply. The article aims to explore the necessity of developing sophisticated multilayer artificial neural networks to compute Chézy’s roughness coefficient.Under the study, the following modifications of the neural network computing Chézy’s roughness coefficient were considered and analysed: (1) Application of two hidden layers of neurons; (2) Application of three hidden layers of neurons; (3) Use of a dropout algorithm for training neural networks by randomly dropping units during training to prevent their co-adaptation; (4) Apart from the sigmoid (logistic) activation function, the use of other artificial neuron transfer functions – hyperbolic tangent (tanh) and rectifying activation function (ReLU).The training and testing of the considered neural network options were carried out using the actual hydro-morphological and hydrological data related to the channel section on the Dnieper River (downstream of Kyiv), the Desna River section near Chernihiv, and the Pripyat River section near the town of Turiv. The Python object-oriented programming environment was applied to build and train the neural networks. The test results confirm the acceptability and sufficiency of computing the Chézy roughness coefficient using the ANN of direct propagation with one hidden layer and a sigmoid logistic activation function. The formation of a qualitative set of training data, as well as data arrangement and choosing a relevant computing model based on empirical knowledge, are, as concluded, among more actual issues than creating more sophisticated neural networks.

Neural network methods for vowel classification in the vocalic systems with the [ATR] (Advanced Tongue Root) contrast

The paper aims to discuss the results of testing a neural network which classifies the vowels of the vocalic system with the [ATR] (Advanced Tongue Root) contrast based on the data of Akebu (Kwa family). The acoustic nature of the [ATR] feature is yet understudied. The only reliable acoustic correlate of [ATR] is the magnitude of the first formant (F1) which can be also modulated by tongue height, resulting in significant overlap between high [-ATR] vowels and mid [+ATR] vowels. Other acoustic metrics which had been associated with the [ATR], such as F1 bandwidth (B1), relative intensity of F1 to F2 (A1-A2), etc., are typically inconsistent across vowel types and speakers. The values of four metrics – F1, F2, A1-A2, B1 – were used for training and testing the neural network. We tested four versions of the model differing in the presence of the fifth variable encoding the speaker and the number of hidden layers. The models which included the variable encoding the speaker achieved slightly higher accuracy, whereas the precision and recall metrics of the three-layer model were generally higher than those with two hidden layers.

CUSTOMS FRAUD DETECTION USING EXTREMELY BOOSTED NEURAL NETWORK (XBNET)

Noticing the vital role of cross-border trade has made Customs plays a crucial role not only in maintaining supply chain but also in securing government revenue from intentional fraud. Given the huge volume of international trade and limited workforce, Customs across the world must implement efficient and effective risk management. This paper proposes XBNet, an ensemble of tree-based algorithms with deep learning algorithms, to detect fraud in import activity. The strength of XBNet is combining gradient-boosted trees with neural networks where the weights, bias, and loss are adjusted simultaneously with the importance features of each tree in each layer. The object of this study is Import Declaration data from four Customs Offices, and the model is set for each Customs office to capture fraud patterns related to their region. We compared the model with two different parameters and concluded the models with learning rates = 1%, number of hidden layers = 2, activation function = sigmoid, and number of epochs = 100 as the most suitable for the Belawan, Merak, and Makassar and model with number of hidden layers = 2, number of epochs = 50 and other parameters are set the same as the most suitable for Tanjung Emas.

New criteria to select reasonable hyperparameters for kinetic parameter estimation in distributed activation energy model (DAEM) by using neural network

The utilization of biomass is being actively explored to reduce net emissions of CO2. Solid carbon such as biomass is a mixture of various components that are difficult to separate, and it is not easy to evaluate its reactivity. The distributed activation energy model (DAEM) is often used to evaluate the reactivity of solid feedstocks such as biomass and coal. In our previous study, we proposed DAEM-NN, a method to estimate the model parameters using a neural network. In this study, we propose a new criterion Φ for selecting the optimal hyperparameters in DAEM-NN: by minimizing Φ, we can select hyperparameters that maximize both the conversion prediction accuracy and the typical importance of each reaction in a parallel reaction system. To verify the usefulness of the proposed criterion, reaction data were generated by numerical simulation, and the number of hidden layer, and the lower limit of contribution were optimized to minimize Φ. The hyperparameters that minimize Φ reduced the set of kinetic parameters corresponding to less important responses from the DAEM-NN kinetic analysis results.

Classification of DNA Sequence Based on a Non-gradient Algorithm: Pseudoinverse Learners.

This chapter proposes a prototype-based classification approach for analyzing DNA barcodes that uses a spectral representation of DNA sequences and a non-gradient neural network. Biological sequences can be viewed as data components with higher non-fixed dimensions, which correspond to the length of the sequences. Through computational procedures such as one-hot encoding, numerical encoding plays an important role in DNA sequence evaluation (OHE). However, the OHE method has some disadvantages: (1) It does not add any details that could result in an additional predictive variable, and (2) if the variable has many classes, OHE significantly expands the feature space. To address these shortcomings, this chapter proposes a computationally efficient framework for classifying DNA sequences of living organisms in the image domain. A multilayer perceptron trained by a pseudoinverse learning autoencoder (PILAE) algorithm is used in the proposed strategy. The learning control parameters and the number of hidden layers do not have to be specified during the PILAE training process. As a result, the PILAE classifier outperforms other deep neural network (DNN) strategies such as the VGG-16 and Xception models.

River system sediment flow modeling using artificial neural networks

Sediment leads to problems with navigation, agricultural productivity, and water pollution. The study of sediment flow in river reaches, which is a non-linear and complex process, is, thus, essential to addressing these issues. The application of artificial neural networks (ANN) to such problems needs to be investigated. For unsteady flow in a river system, river reach storage is an important variable that needs to be considered in data-driven models. However, previous research on sediment modeling did not involve the explicit use of storage variables in such models as is investigated in the current study. In the current study, storage variables have been explicitly (Model 2) used to predict the output state of the system at time ‘t + 1’ from the input state at time ‘t’ using ANNs. Sediment discharge at six gaging stations on the Mississippi River system, USA, has been considered as the state variable. The model has been compared with a model considering implicit variation of the storage parameter in the river system (Model 1). Dynamic ANNs are used for time-series datasets, which are more suitable for incorporating the sequential information within the dataset. Focussed gamma memory neural networks have been used in the current study. The numbers of hidden layers and hidden nodes, activation function, and learning rate have been varied step by step to obtain the optimal ANN configurations. The best selected input–output variables are those used in Model 2 as it performed slightly better than the other model in terms of Nash–Sutcliffe efficiency coefficient (CE) values. Model performance evaluated using normalized root mean square error (NRMSE) and CE shows satisfactory results. NRMSE was < 10% for all the outputs except for the Venedy and Murphysboro locations and CE values for sediment loads were > 0.45 for all locations except Murphysboro indicating acceptable performance by both the models. The models proved highly efficient (CE > 0.80, i.e., very good predictions) for predicting sediment discharge at locations along the main river channel with acceptable accuracy (CE > 0.45) for other locations and the storage change for the river system. These models can be used for real-time forecasting and management of sediment-related problems.

Particle swarm optimization of Elman neural network applied to battery state of charge and state of health estimation

Lithium-ion batteries have emerged as an energy storage solution for electrified vehicles. A Battery Management System (BMS) is critical for efficient and reliable system operation, in which State of Charge (SoC) estimation and State of Health (SoH) monitoring are of major importance to ensure optimal energy management in battery vehicles for increased autonomy and battery life. This paper presents a neural network with Elman architecture trained for a lithium-ion cell, aiming at SoC and SoH estimation. The multi-objective optimization approach based on the particle swarm algorithm is used for the training in order to lower the root mean square error in calculating the SoC and SoH. For such purposes, the neural network characteristics are optimized, such as the number of hidden layers, the number of neurons in each hidden layer, the activation functions, the bias value, and the weights of the inputs and outputs. The best trade-off solution has an error of 2.56% in the average SoC estimate and 0.003% in the average SoH estimate.

Improving the efficiency and stability of perovskite solar cell: Application of innovative machine learning algorithm

Solar cells are modern inventions that use the photovoltaic effect to directly convert light energy into electricity, generating electrical charges that are free to move through semiconductors. The semiconductor is typically utilized as the raw material for solar cells. In order to convert energy, electron–hole pairs that are responsible for producing light (photon) energy must be absorbed in a semiconductor, followed by charge carrier separation. Enhancing solar cells’ stability is crucial in engineering because they are used in a variety of environments. Moreover, graphene nanoplatelets (GPLs) have a great deal of potential to enhance ceramic–GNP composites’ mechanical, tribological, electrical, thermal, and biological characteristics, all at once. Machine learning algorithms (MLA) are often used to forecast how various systems would behave. In an MLA network, hyperparameters like the number of hidden layers and learning rate are often selected manually as required. MLA is used in this work to examine spinning cylindrical constructions’ stability at the microscale. In this context, the particle swarm optimization (PSO) is used to optimize the weights and biases of the network. The number of perceptions in the two hidden layers is optimized in a second parallel process using a genetic algorithm. The modified torque–stress theory (MCST) equation’s numerical solution for the dynamic behavior of GPL-reinforced perovskite solar cells was used to train the MLA. It is proposed to guide the spatial discretization of governing equations using the variational differential quadrature (VDQ) method as a direct discretization of the energy functional in the space domain. Lastly, the findings demonstrate that the stability of the current cantilevered solar cell reinforced by GPLs is significantly influenced by curvature, length scale, the shape of the solar cell, and mode number factors.

Building an IoT temperature and humidity forecasting model based on long short-term memory (LSTM) with improved whale optimization algorithm

In particular, predicting the temperature and humidity information plays a crucial role in plantation, estimating rainfalls and climate change, and predicting air quality via specified geographical regions. The temperature and humidity forecasting information is occasionally presented with low accuracy due to uncertain techniques and vast methods that employ different sensors and models. For this reason, this work proposes an Internet of Things (IoT) temperature and humidity forecasting model based on an improved whale optimization algorithm with long short-term memory (IWOA-LSTM) technique. To increase the convergence speed processing time and overcome the local optimization problem, the IWOA is introduced. The number of hidden layers, learning rate momentum, and weight decay of the LSTM optimized using the IWOA. The actual temperature and humidity data are collected using DHT11 and ESP8266 NodeMCU practical model and processed using the ThingSpeak platform. The processing data stage depends on filling the missing data gaps using the rolling average technique (RAT). The performance evaluation of the proposed IWOA-LSTM forecasting model is assessed using some statistical functions, namely known as mean square error, mean absolute error, root mean square error, and mean absolute percentage error. The IWOA-LSTM techniques were also assessed using throughput, latency, and power consumption. The developed IWOA-LSTM model shows high accuracy, leading to better forecasting information than other forecasting models.

Deep learning for plant bioinformatics: an explainable gradient-based approach for disease detection.

Emerging in the realm of bioinformatics, plant bioinformatics integrates computational and statistical methods to study plant genomes, transcriptomes, and proteomes. With the introduction of high-throughput sequencing technologies and other omics data, the demand for automated methods to analyze and interpret these data has increased. We propose a novel explainable gradient-based approach EG-CNN model for both omics data and hyperspectral images to predict the type of attack on plants in this study. We gathered gene expression, metabolite, and hyperspectral image data from plants afflicted with four prevalent diseases: powdery mildew, rust, leaf spot, and blight. Our proposed EG-CNN model employs a combination of these omics data to learn crucial plant disease detection characteristics. We trained our model with multiple hyperparameters, such as the learning rate, number of hidden layers, and dropout rate, and attained a test set accuracy of 95.5%. We also conducted a sensitivity analysis to determine the model's resistance to hyperparameter variations. Our analysis revealed that our model exhibited a notable degree of resilience in the face of these variations, resulting in only marginal changes in performance. Furthermore, we conducted a comparative examination of the time efficiency of our EG-CNN model in relation to baseline models, including SVM, Random Forest, and Logistic Regression. Although our model necessitates additional time for training and validation due to its intricate architecture, it demonstrates a faster testing time per sample, offering potential advantages in real-world scenarios where speed is paramount. To gain insights into the internal representations of our EG-CNN model, we employed saliency maps for a qualitative analysis. This visualization approach allowed us to ascertain that our model effectively captures crucial aspects of plant disease, encompassing alterations in gene expression, metabolite levels, and spectral discrepancies within plant tissues. Leveraging omics data and hyperspectral images, this study underscores the potential of deep learning methods in the realm of plant disease detection. The proposed EG-CNN model exhibited impressive accuracy and displayed a remarkable degree of insensitivity to hyperparameter variations, which holds promise for future plant bioinformatics applications.