Sigmoid Activation Function Research Articles

Edge segmentation method for Si3N4 bearing rolling elements microcracks with profile-distortion

For the profile-distortion of Si3N4 bearing rolling elements microcracks image. A coupling method of variable-scale truncated median filtering and detection-driven active contour model is proposed. Details are as follows. The sigmoid activation function is constructed to dynamically adjust the circular window. The local threshold truncation is designed to adapt to the characteristics of different regions in the image. The area threshold is defined to eliminate non-target features. The mean peak signal-to-noise ratio (PSNR) of filtered images is about 40.263. The edge overlap rate is as low as 0.1027. The mean accuracy of feature extraction is about 94.33 %. To sum up, the influence result from profile-distortion on the accuracy of feature extraction of Si3N4 bearing rolling elements microcracks is effectively solved.

Exponential Sampling Type Kantorovich Max-Product Neural Network Operators

This article deals with the study of approximation behavior of the exponential sampling type Kantorovich max-product neural network operators based on sigmoidal activation function. We study point-wise and uniform convergence theorems in the space of continuous functions for these operators and determine the rate of convergence by means of logarithmic modulus of continuity. We also investigate the approximation of p -th ( 1 ≤ p < ∞ ) Lebesgue integrable functions by the aforementioned operators and study the rate of convergence exploiting Peetre’s K − functional.

Early detection of serious but struggling learners in MOOCs using ensemble deep learning

ABSTRACT This study works on the early identification of serious but struggling students in Massive Open Online Courses (MOOCs), which facilitates a transition to a hybrid learning format. Firstly, we exclusively utilize basic behavior log features that are universally available across all MOOC platforms. Secondly, we encode each learning activity with a distinct identifier, which helps us utilize the current advances in NLP to address our problem. Thirdly, we seamlessly integrate our LSTM model with a Logistic Regression model, configured as a single dense layer with a sigmoid activation function, within the broader framework of deep learning.

The Application of an Improved LESS Dung Beetle Optimization in the Intelligent Topological Reconfiguration of ShipPower Systems

To address the shortcomings of the Dung Beetle Optimization (DBO) algorithm in ship power-system fault reconfiguration, such as low population diversity and an imbalance between global exploration and local exploitation, the authors of this paper propose an improved Dung Beetle Optimization (LESSDBO) algorithm. The improvements include optimizing the initial population using Latin hypercube sampling and an elite population strategy, optimizing parameters with an improved sigmoid activation function, introducing the sine–cosine algorithm (SCA) for position update optimization, and performing multi-population mutation operations based on individual quality. The LESSDBO algorithm was applied to simulate the fault reconfiguration of a ship power system, and it was compared with the traditional DBO, Genetic Algorithm (GA), and Modified Particle Swarm Optimization (MSCPSO) methods. The simulation results showed that LESSDBO outperformed the other algorithms in terms of convergence accuracy, convergence speed, and global search capability. Specifically, in the reconfiguration under Fault 1, LESSDBO achieved optimal convergence in seven iterations, reducing convergence iterations by more than 30% compared with the other algorithms. In the reconfiguration under Fault 2, LESSDBO achieved optimal convergence in eight iterations, reducing convergence iterations by more than 23% compared with the other algorithms. Additionally, in the reconfiguration under Fault Condition 1, LESSDBO achieved a minimum of four switch actions, which is 33% fewer than the other algorithms, on average. In the reconfiguration under Fault Condition 2, LESSDBO achieved a minimum of eight switch actions, which is a 5.9% reduction compared with the other algorithms. Furthermore, LESSDBO obtained the optimal reconfiguration solution in all 50 trials for both Faults 1 and 2, demonstrating a 100% optimal convergence probability and significantly enhancing the reliability and stability of the algorithm. The proposed method effectively overcomes the limitations of the traditional DBO in fault reconfiguration, providing an efficient and stable solution for the intelligent topology reconfiguration of ship power systems.

Deep neural networks and conventional machine learning classifiers to analyze thoracic survival data

Lung cancer is a prevalent global health concern and most prevalent malignancy in Indonesian hospitals. Following thoracic surgery, patients were categorized into two classes: individuals who experienced mortality within a year and those who achieved survival. Despite being about socks, the dataset for the deceased category consisted of 70 data samples, while the dataset for the final group comprised 400 samples. Data calculation involves the utilization of both deep neural networks and standard machine learning algorithms. The study use the Python programming language to evaluate the algorithms, and it measures their performance using metrics such as accuracy, F1-Score, precision, recall, receiver operating characteristic (ROC), and area under curve (AUC). The test results indicate that the deep neural network method achieves an accuracy of 95,56%, an F1 score of 79,24%, a precision of 91,96%, a recall of 85,52%, and an AUC of 85,52%. This study suggests that utilizing deep neural network data mining techniques, specifically with a cross-validation fold of 10, variations of six hidden layer encoder-decoder, relu, sigmoid activation function, optimizer Adam, and learning rate of 0,01, dropout rate of 0,2. Employing the Synthetic Minority Over-sampling Technique data preprocessing method, can effectively analyze thoracic patient survival data sets.

PERAMALAN INDEKS HARGA SAHAM GABUNGAN (IHSG) MENGGUNAKAN METODE BACKPROPAGATION DAN RECURRENT NEURAL NETWORK (RNN)

Forecasting is an objective calculation and uses past data to predict something in the future. The economic activity of buying and selling shares is a common activity in this digital era. Investment in the capital market is currently popular among the general public. Forecasting in the digitalization era has undergone many developments, as well as its application not only in agriculture, but also in economics, politics and other fields. The development of forecasting in the economic field is growing rapidly with the use of more complex mathematical models and the application of information technology. Forecasting is an objective calculation and uses past data to predict something in the future. Utilization of artificial neural network (ANN) is one of the solutions for forecasting with more accurate results. One of the algorithms in ANN is the backpropagation algorithm. Apart from using other backpropagation forecasting methods that can be used is the deep learning (DL) method with a recurrent neural network algorithm. The backpropagation method produces an optimal MAPE of 13.28%. with an architectural model of seven input layer neurons, seven hidden layer neurons, and using the sigmoid activation function. The recurrent neural network method produces an optimal MAPE of 1.12%. with an architectural model of seven input layer neurons, ten hidden layer neurons, and using the tanh activation function. If the two optimal MAPEs of each method are compared, the Recurrent Neural Network Method produces more optimal forecasting compared to the backpropagation method.

Skin Cancer Classification Using Fine-Tuned Transfer Learning of DENSENET-121

Skin cancer diagnosis greatly benefits from advanced machine learning techniques, particularly fine-tuned deep learning models. In our research, we explored the impact of traditional machine learning and fine-tuned deep learning approaches on prediction accuracy. Our findings reveal significant improvements in predictability and accuracy with fine-tuning, particularly evident in deep learning models. The CNN, SVM, and Random Forest Classifier achieved high accuracy. However, fine-tuned deep learning models such as EfficientNetB0, ResNet34, VGG16, Inception _v3, and DenseNet121 demonstrated superior performance. To ensure comparability, we fine-tuned these models by incorporating additional layers, including one flatten layer and three densely interconnected layers. These layers play a crucial role in enhancing model efficiency and performance. The flatten layer preprocesses multidimensional feature maps, facilitating efficient information flow, while subsequent dense layers refine feature representations, capturing intricate patterns and relationships within the data. Leveraging LeakyReLU activation functions in the dense layers mitigates the vanishing gradient problem and promotes stable training. Finally, the output dense layer with a sigmoid activation function simplifies decision making for healthcare professionals by providing binary classification output. Our study underscores the significance of incorporating additional layers in fine-tuned neural network models for skin cancer classification, offering improved accuracy and reliability in diagnosis.

Schizophrenia diagnosis using the GRU-layer's alpha-EEG rhythm's dependability

Verifying schizophrenia (SZ) can be assisted by deep learning techniques and patterns in brain activity observed in alpha-EEG recordings. The suggested research provides evidence of the reliability of alpha-EEG rhythm in a Gated-Recurrent-Unit-based deep-learning model for investigating SZ. This study suggests Rudiment Densely-Coupled Convolutional Gated Recurrent Unit (RDCGRU) for the various EEG-rhythm-based (gamma, beta, alpha, theta, and delta) diagnoses of SZ. The model includes multiple 1-D-Convolution (Con-1-D) folds with steps greater than 1, which enables the model to programmatically and effectively learn how to reduce the incoming signal. The Con-1-D layers and numerous Gated Recurrent Unit (GRU) layers comprise the Exponential-Linear-Unit activation function. This powerful activation function facilitates in-deep-network training and improves classification performance. The Densely-Coupled Convolutional Gated Recurrent Unit (DCGRU) layers enable RDCGRU to address the training accuracy loss brought on by vanishing or exploding gradients, and this might make it possible to develop intense, deep versions of RDCGRU for more complex problems. The sigmoid activation function is implemented in the digital (binary) classifier's output nodes. The RDCGRU deep learning model attained the most excellent accuracy, 88.88 %, with alpha-EEG rhythm. The research achievements: The RDCGRU deep learning model's GRU cells responded superiorly to the alpha-EEG rhythm in EEG-based verification of SZ.

Improving Stock Price Forecasting Accuracy with Stochastic Multilayer Perceptron

The stock market operates in a stochastic environment, making accurate price forecasting challenging. To address this issue, a stochastic multilayer perceptron (S-MLP) model has been developed to simulate the stock market's stochastic nature. By incorporating a Gaussian process into the sigmoid activation function, this model incorporates stochasticity into the traditional multilayer perceptron (MLP). As the perturbation factor, a stochastic sigmoid activation function (SAF) with a volatility estimator is used. Although S-MLP has demonstrated superiority over MLP, there is still room for improvement in terms of forecasting precision. In this study, we propose S-MLP with a trainable perturbation factor (S-MLPT), an improved variant of S-MLP. SAF employs the Yang-Zhang volatility estimator as the perturbation factor. The proposed model first employed MLP, and all the parameters were trained. After freezing the parameters, S-MLP is used to train the perturbation factor in the SAF. To evaluate the predictive performance of the models, MLP, S-MLP, and S-MLPT are used to predict the one day ahead highest stock price of four counters listed in Bursa Malaysia. As an evaluation metric, the coefficient of determination is utilised, and the relative percentage improvement of the models is calculated to determine their superiority. The results demonstrated that S-MLP outperforms MLP by effectively minimizing the loss function and converging towards a better local or global minimum during training. In conclusion, S-MLPT exhibits even better performance than S-MLP, with relative percentage improvements of 0.14%, 15.45%, and 0.48% for counters 0166.KL, 2445.KL, and 4707.KL, respectively.

A 3D Convolutional Neural Network Based on Non-enhanced Brain CT to Identify Patients with Brain Metastases.

Dedicated brain imaging for cancer patients is seldom recommended in the absence of symptoms. There is increasing availability of non-enhanced CT (NE-CT) of the brain, mainly owing to a wider utilization of Positron Emission Tomography-CT (PET-CT) in cancer staging. Brain metastases (BM) are often hard to diagnose on NE-CT. This work aims to develop a 3D Convolutional Neural Network (3D-CNN) based on brain NE-CT to distinguish patients with and without BM. We retrospectively included NE-CT scans for 100 patients with single or multiple BM and 100 patients without brain imaging abnormalities. Patients whose largest lesion was < 5mm were excluded. The largest tumor was manually segmented on a matched contrast-enhanced T1 weighted Magnetic Resonance Imaging (MRI), and shape radiomics were extracted to determine the size and volume of the lesion. The brain was automatically segmented, and masked images were normalized and resampled. The dataset was split into training (70%) and validation (30%) sets. Multiple versions of a 3D-CNN were developed, and the best model was selected based on accuracy (ACC) on the validation set. The median largest tumor Maximum-3D-Diameter was 2.29cm, and its median volume was 2.81cc. Solitary BM were found in 27% of the patients, while 49% had > 5 BMs. The best model consisted of 4 convolutional layers with 3D average pooling layers, dropout layers of 50%, and a sigmoid activation function. Mean validation ACC was 0.983 (SD: 0.020) and mean area under receiver-operating characteristic curve was 0.983 (SD: 0.023). Sensitivity was 0.983 (SD: 0.020). We developed an accurate 3D-CNN based on brain NE-CT to differentiate between patients with and without BM. The model merits further external validation.

Physics-informed neural network classification framework for reliability analysis

Reliability analysis is crucial for quantitatively evaluating structural safety amidst uncertainties, laying the foundation for reliability-based design optimization aimed at augmenting structural reliability and reducing economic expenditures, thereby offering substantial engineering benefits. However, performing reliability analysis on intricate structures, especially those requiring time-intensive finite element models, presents a formidable challenge. This study develops an innovative physics-informed neural network classification (PINNC) model to tackle this issue. In the PINNC model, the loss associated with the structural output state (i.e., safety or failed state) is defined as classification loss. Additionally, the structural output value (i.e., actual structural response) is treated as critical physical information, with its associated loss termed as physical loss. To facilitate a separate calculation of physical loss and classification loss, a parametric sigmoid activation function is used, establishing a link between the structural output value and the structural output state. The total loss is calculated as a weighted sum of both physical and classification losses. Unlike conventional neural network classification models that solely focus on classification loss, the PINNC model integrates both physical and classification losses, markedly improving the model’s efficacy in structural reliability analysis. Moreover, an adaptive framework is developed to incrementally include samples proximal to the limit state surface as new training data, thereby reinforcing the PINNC model’s computational precision. The effectiveness of the proposed PINNC framework in overcoming reliability analysis challenges is demonstrated through various applications.

Enhancing liver tumor segmentation with UNet-ResNet: Leveraging ResNet’s power

BACKGROUND: Liver cancer poses a significant health challenge due to its high incidence rates and complexities in detection and treatment. Accurate segmentation of liver tumors using medical imaging plays a crucial role in early diagnosis and treatment planning. OBJECTIVE: This study proposes a novel approach combining U-Net and ResNet architectures with the Adam optimizer and sigmoid activation function. The method leverages ResNet’s deep residual learning to address training issues in deep neural networks. At the same time, U-Net’s structure facilitates capturing local and global contextual information essential for precise tumor characterization. The model aims to enhance segmentation accuracy by effectively capturing intricate tumor features and contextual details by integrating these architectures. The Adam optimizer expedites model convergence by dynamically adjusting the learning rate based on gradient statistics during training. METHODS: To validate the effectiveness of the proposed approach, segmentation experiments are conducted on a diverse dataset comprising 130 CT scans of liver cancers. Furthermore, a state-of-the-art fusion strategy is introduced, combining the robust feature learning capabilities of the UNet-ResNet classifier with Snake-based Level Set Segmentation. RESULTS: Experimental results demonstrate impressive performance metrics, including an accuracy of 0.98 and a minimal loss of 0.10, underscoring the efficacy of the proposed methodology in liver cancer segmentation. CONCLUSION: This fusion approach effectively delineates complex and diffuse tumor shapes, significantly reducing errors.

Application of Artificial Neural Network Analysis in Predicting the Performance of Microbial Energy Cells

Objective: The objective of this study was to apply Artificial Neural Networks to evaluate the performance of Microbial Energy Cells, to identify the best network configuration for cell evaluation. Theoretical Framework: Although several of the widely used effluent treatment methods show results, most of them have a common disadvantage: they lose the chemical energy contained in the treated effluent and have high energy consumption for their conduction. Therefore, an increasing effort has been made to develop effluent treatment technologies capable of recovering part of the energy contained in the waste to be treated. In this scenario, microbial energy cells (CEM) emerge as a potential technology, as they are devices that simultaneously treat effluent biologically and generate electrical energy. Methodology: For the application and evaluation of ANNs in CEM, a feedforward neural network was used, with a Levenberg-Marquardt training algorithm, 1 or 2 hidden layers, with sigmoid and tansig activation functions, and an accuracy factor of 10-5. The data used for training and validation for the ANN were obtained through a literature search. Networks with 15, 30, 50, 90, 100, 130, 150, and 200 neurons were used for testing to evaluate the best performance. Results and Discussion: With the results obtained, it was observed that the best adjustment of the network occurred with the 2-layer configuration, one layer with 100 neurons and the other output layer, with 49 interactions and R2 of 0.91 in the training adjustment, 0 .78 in the validation fit and 0.90 in the fit with all experimental data evaluated, respectively. Originality/Value: This study contributes to the literature by evaluating the application of artificial neural networks, which are empirical modeling mechanisms, inspired by biological nervous systems, with processing abilities, in microbial energy cells.

A neural network computational procedure for the novel designed singular fifth order nonlinear system of multi-pantograph differential equations

The current investigations present the numerical solutions of the novel singular nonlinear fifth-order (SNFO) system of multi-pantograph differential model (SMPDM), i.e., SNFO–SMPDM. The novel SNFO–SMPDM is obtained using the sense of the second kind of typical Emden–Fowler and prediction differential models. The features of shape factor, pantograph along with singular points are provided for all four obtained classes of the SNFO–SMPDM. The extensive use of the singular models is observed in the engineering and mathematical systems, e.g., inverse systems and viscoelasticity or creep systems. For the correctness of the proposed novel SNFO–SMPDM, one case of each class is numerically handled by applying supervised neural networks (SNNs) along with the optimization of Levenberg–Marquardt backpropagation scheme (LMBS), i.e., SNNs–LMBS. A dataset using the traditional variational iteration scheme is designed to compare the proposed results of each case of SNFO–SMPDM. The obtained approximate solutions of each class using the novel SNFO-SMPDM are presented based on the training (80 %), authentication (10 %) and testing (10 %) measures to evaluate the mean square error. Fifteen numbers of neurons, and sigmoid activation function are used in this SNN process. To authenticate the competence, and precision of SNFO–SMPDM, the numerical simulations are accessible by applying the relative measures of regression, error histogram plots, and correlation.

An attention-based deep learning for acute lymphoblastic leukemia classification

The bone marrow overproduces immature cells in the malignancy known as Acute Lymphoblastic Leukemia (ALL). In the United States, about 6500 occurrences of ALL are diagnosed each year in both children and adults, comprising nearly 25% of pediatric cancer cases. Recently, many computer-assisted diagnosis (CAD) systems have been proposed to aid hematologists in reducing workload, providing correct results, and managing enormous volumes of data. Traditional CAD systems rely on hematologists’ expertise, specialized features, and subject knowledge. Utilizing early detection of ALL can aid radiologists and doctors in making medical decisions. In this study, Deep Dilated Residual Convolutional Neural Network (DDRNet) is presented for the classification of blood cell images, focusing on eosinophils, lymphocytes, monocytes, and neutrophils. To tackle challenges like vanishing gradients and enhance feature extraction, the model incorporates Deep Residual Dilated Blocks (DRDB) for faster convergence. Conventional residual blocks are strategically placed between layers to preserve original information and extract general feature maps. Global and Local Feature Enhancement Blocks (GLFEB) balance weak contributions from shallow layers for improved feature normalization. The global feature from the initial convolution layer, when combined with GLFEB-processed features, reinforces classification representations. The Tanh function introduces non-linearity. A Channel and Spatial Attention Block (CSAB) is integrated into the neural network to emphasize or minimize specific feature channels, while fully connected layers transform the data. The use of a sigmoid activation function concentrates on relevant features for multiclass lymphoblastic leukemia classification The model was analyzed with Kaggle dataset (16,249 images) categorized into four classes, with a training and testing ratio of 80:20. Experimental results showed that DRDB, GLFEB and CSAB blocks’ feature discrimination ability boosted the DDRNet model F1 score to 0.96 with minimal computational complexity and optimum classification accuracy of 99.86% and 91.98% for training and testing data. The DDRNet model stands out from existing methods due to its high testing accuracy of 91.98%, F1 score of 0.96, minimal computational complexity, and enhanced feature discrimination ability. The strategic combination of these blocks (DRDB, GLFEB, and CSAB) are designed to address specific challenges in the classification process, leading to improved discrimination of features crucial for accurate multi-class blood cell image identification. Their effective integration within the model contributes to the superior performance of DDRNet.

DAILY DANUBE RIVER WATER LEVEL PREDICTION USING EXTREME LEARNING MACHINE APPROACH

Anticipating water levels in vast riverbeds is crucial for preventing and mitigating floods or droughts, assessing power plant capacity, and facilitating navigation management. This study introduces an innovative water level prediction model utilizing an Extreme Learning Machine developed to solve the issues of low performance of existing forecasting methods. Development of such a system is of extreme importance when talking about the largest European river – the Danube River. Experimental findings reveal the model's satisfactory performance across various accuracy metrics, complexity considerations, and calculation speed. The prediction with the highest error rate based on MAPE criteria was for Prahovo water level prediction over a 365-day period at 2.02%, whilst the most accurate predictions were for Novi Sad and Banatska Palanka over 30 days and 180 days horizons, respectively, at 0.0550%. The highest coefficient of determination (R2) was achieved with the Novi Sad data at 0.9968, whilst the lowest was observed with the Prahovo data at 0.7353. The ELM model achieved high precision by adjusting the activation functions of the hidden layer neurons, which involved using different combinations of sigmoid and radial-basis activation functions.

Heterogeneous wireless network selection using feed forward double hierarchy linguistic neural network

Network selection in heterogeneous wireless networks (HWNs) is a complex issue that requires a thorough understanding of service features and user preferences. This is because the various wireless access technologies have varying capabilities and limitations, and the best network for a voice, video, and data service depends on a variety of factors. For selecting the optimal network in HWNs, varying factors such as the user’s position, accessible network resources, service quality requirements, and user preferences must be considered. The classical decision making procedure is very difficult and uncertain to select the desirable HWNs for voice, video, and data. Therefore, we develop a novel decision making model based on feed-forward neural networks under the double hierarchy linguistic information for the selection of the best HWNs for voice, video, and data. In this article, we introduce a novel feed-forward double hierarchy linguistic neural network using the Hamacher t-norm and t-conorm. Further, the feed-forward double hierarchy linguistic neural network applies to the decision making model for the selection of the best HWNs for voice, video, and data. In this novel decision making model, we first take the given data about HWNs and use the converting function to convert the given data into a double hierarchy linguistic term set. We calculate the hidden layer and output layer information by using Hamacher aggregation operations. Finally, we use the sigmoid activation function on the output layer information to decide on the best HWNs for voice, video, and data according to ranking. The proposed approach is compared with other existing models of decision making and the results of the comparison show that the proposed technique is applicable and reliable for the decision support model.

Flame propagation speed prediction model of premixed methane gas deflagration experiments based on Adamax-LSTM for FLNG

A time-series prediction method based on AdaMax-LSTM neural network is proposed for predicting the flame propagation speed in premixed methane gas deflagration experiments, which can provide a decision-making basis for emergency operation of FLNG combustible gas deflagration accidents. Firstly, 54 sets of premixed methane gas deflagration experiments under semi-open duct obstacle conditions were conducted to investigate the different deflagration mechanisms by changing the obstacle parameters. The experimental results demonstrate that the distance between the obstacle and ignition source, obstacle length and obstacle shape will all effect the flame propagation speed and deflagration overpressure. Secondly, the LSTM neural network is employed to setup a novel method which can predict the flame speed in time series via calculating the Reynolds number and determining the turbulence of the flame accurately. The deflagration experiments results were used as the dataset for AI training for the proposed prediction method. In addition, the AdaMax optimizer is added into the backpropagation process of the proposed LSTM neural network to maximize the prediction accuracy of the method. The analysis results indicate that the AdaMax-LSTM neural network with sigmoid activation function can achieve the highest level of accuracy prediction, with the mean R2 value reaching 0.95, and can identify anomaly data and the most different deflagration mechanisms experimental condition. The proposed method provides an efficient and accurate way to predict and analyze the deflagration mechanisms via employing cutting-edge AI technology.

Comparative Analysis of Recurrent Neural Network Models Performance in Predicting Bitcoin Prices

The recurring neural network is a deep learning algorithm that is commonly used to develop prediction systems. There are many variants of RNN such as RNN itself, long-short-term memory (LSTM), and gated recurring unit, so it is frequently debatable which algorithm from the RNN family has the most optimal efficiency and computation time. When developing a prediction system, sequential or time series data is required so that an accurate prediction can be made. Sequential or time series data involve data arranged in a time sequence, such as weather data, financial data, carbon emission data, and traffic data recorded over time. This research will be carried out by predicting the three RNN models against historical Bitcoin value data. The research method used is Experimental Design by comparing the performance between the three models on bitcoin value time series data, testing is done by involving hyperparameters such as Tanh, Sigmoid, and ReLU activation functions, batch size, and epochs. The aim of this research is to find out which RNN model can produce the most optimal performance and find out what performance measures can be used to evaluate and compare the performance between the three models. The results of the study show that LSTM is the most effective model with RMSE 0.012441 and MSE 0.000155 but inefficient because it takes 3 minutes 24 seconds to run the computation; in the meantime, the Tanh activation function gives the most optimal prediction than Sigmoid and RelU and therefore should be the main candidate to be used with RNN models when predicting Bitcoin prices.

Modelling thermal characteristics of cocoa butter using a feed‐forward artificial neural network based on multilayer perceptron

SummaryCocoa butter is the most important ingredient of chocolate, which determines its melting behaviour. Variations in the melting characteristics of cocoa butter can profoundly affect the performance and suitability for their industrial utilisation. Over time, researchers have been attempting to establish a logical relationship between cocoa butter's unique thermal properties and the amount of saturated to unsaturated fatty acids in mono, di and triglycerides, and fatty acids (as major components), and free fatty acids, soap, primary oxidation products, minerals, moisture, phospholipids, tocopherols, unsaponifiable matters and metals (as minor components) found in cocoa butter. In this research, the thermal behaviours of thirteen samples of cocoa butter with different origins were investigated using isothermal differential scanning calorimetry. The cocoa butters starting temperature of crystallisation, temperature of maximum heat release, temperature of completed crystallisation and the enthalpy of heat release during recrystallisation were evaluated. In addition, the chemical composition (moisture, acidity, peroxide, minerals and soap content), fatty acid and triacylglycerol composition were used to establish an MLP‐ANN with fourteen input neurons connected by two flexible, sigmoid activation function layers. The back‐propagation was used to train the artificial neural network (ANN) structure and optimise the error of prediction. The study showed that the MLP algorithm can predict the thermal behaviour of CB samples with trace error, regardless of plant growth and extract process condition.