Average Prediction Accuracy Research Articles

Fire Detection by using DenseNet 201 algorithm and Surveillance Cameras images

Early warning of fires is very important to reduce loss of life and property. It is a technology that has become one of the most important design elements for smart cities in the future. Most current solutions rely on temperature and smoke sensors, which have limited detection ranges and scenarios and long response times. These are used inside some buildings and cannot be used in the external environment, as they are not a preferred solution. In addition to the above, sensor systems are costly because of the hardware and installation requirements, and they require regular upkeep. With the development of artificial intelligence and learning machines, fire detection has been used based on deep learning. Widely. Therefore, a solution to computer vision technology is proposed, which uses a convolutional neural network (CNN), Reducing false alarms. The use of computer vision techniques is certainly better than sensor-based systems, by utilizing security camera footage at specific periods to detect a fire when it breaks out. Our system achieved excellent results with an average predictive accuracy of 98% on the dataset. This accuracy and low cost make it a good alternative to systems that use sensors. The contribution of this study is to improve the detection process by conducting a classification process instead of detecting an object (fire), which requires complex mathematical calculations, large resources, and a long time.

Quantification of caffeine in coffee cans using electrochemical measurements, machine learning, and boron-doped diamond electrodes.

Electrochemical measurements, which exhibit high accuracy and sensitivity under low contamination, controlled electrolyte concentration, and pH conditions, have been used in determining various compounds. The electrochemical quantification capability decreases with an increase in the complexity of the measurement object. Therefore, solvent pretreatment and electrolyte addition are crucial in performing electrochemical measurements of specific compounds directly from beverages owing to the poor measurement quality caused by unspecified noise signals from foreign substances and unstable electrolyte concentrations. To prevent such signal disturbances from affecting quantitative analysis, spectral data of voltage-current values from electrochemical measurements must be used for principal component analysis (PCA). Moreover, this method enables highly accurate quantification even though numerical data alone are challenging to analyze. This study utilized boron-doped diamond (BDD) single-chip electrochemical detection to quantify caffeine content in commercial beverages without dilution. By applying PCA, we integrated electrochemical signals with known caffeine contents and subsequently utilized principal component regression to predict the caffeine content in unknown beverages. Consequently, we addressed existing research problems, such as the high quantification cost and the long measurement time required to obtain results after quantification. The average prediction accuracy was 93.8% compared to the actual content values. Electrochemical measurements are helpful in medical care and indirectly support our lives.

Genotype, environment, and their interaction effects on peanut seed protein, oil, and fatty acid content variability

AbstractProducts from peanut seeds are nutritious for human and/or animal consumption. Peanut seed nutritional quality is determined by the genotype (G) of cultivars or accessions, environmental conditions (E), and their interactions (G × E). To evaluate the effects of genotype, environment, and their interactions on seed nutritional quality, 52 peanut germplasm accessions, which vary in oil content and fatty acid composition, were planted at three locations (Citra, FL; Byron, GA; and Clovis, NM) for 2 years (2017 and 2018). The harvested seeds were analyzed for protein, oil, and fatty acid composition using an N analyzer, nuclear magnetic resonance, and gas chromatography, respectively. Significant effects of accession, FAD2 gene, where FAD is fatty acid desaturase, location, year, and their interactions on the investigated traits were evaluated. The average oil content from accessions grown in Florida (51.1%) or Georgia (50.7%) was significantly higher than those grown in New Mexico (45.8%), whereas the average protein content from accessions grown in Florida (23.2%) was significantly lower than those grown in New Mexico (25.4%) or Georgia (24.8%). This is expected as there is a negative correlation between oil and protein content. After genotyping and classification (A/A, A/W, and G/W) for the FAD2A/FAD2B genes, the nutritional quality variations were further evaluated at the levels of the specific haplotype of FAD2A/FAD2B and explicit environmental index. The average prediction accuracy of seed nutritional quality trait values from the prediction model demonstrated that predicting trait values in new environments is feasible. The results of this study provide insights into the seed nutritional analysis, germplasm evaluation, and peanut breeding, cultivation, and production to peanut breeders, curators, farmers, and product processors.

Performance Differences of a Touch-Based Serial Reaction Time Task in Healthy Older Participants and Older Participants With Cognitive Impairment on a Tablet: Experimental Study.

Digital neuropsychological tools for diagnosing neurodegenerative diseases in the older population are becoming more relevant and widely adopted because of their diagnostic capabilities. In this context, explicit memory is mainly examined. The assessment of implicit memory occurs to a lesser extent. A common measure for this assessment is the serial reaction time task (SRTT). This study aims to develop and empirically test a digital tablet-based SRTT in older participants with cognitive impairment (CoI) and healthy control (HC) participants. On the basis of the parameters of response accuracy, reaction time, and learning curve, we measure implicit learning and compare the HC and CoI groups. A total of 45 individuals (n=27, 60% HCs and n=18, 40% participants with CoI-diagnosed by an interdisciplinary team) completed a tablet-based SRTT. They were presented with 4 blocks of stimuli in sequence and a fifth block that consisted of stimuli appearing in random order. Statistical and machine learning modeling approaches were used to investigate how healthy individuals and individuals with CoI differed in their task performance and implicit learning. Linear mixed-effects models showed that individuals with CoI had significantly higher error rates (b=-3.64, SE 0.86; z=-4.25; P<.001); higher reaction times (F1,41=22.32; P<.001); and lower implicit learning, measured via the response increase between sequence blocks and the random block (β=-0.34; SE 0.12; t=-2.81; P=.007). Furthermore, machine learning models based on these findings were able to reliably and accurately predict whether an individual was in the HC or CoI group, with an average prediction accuracy of 77.13% (95% CI 74.67%-81.33%). Our results showed that the HC and CoI groups differed substantially in their performance in the SRTT. This highlights the promising potential of implicit learning paradigms in the detection of CoI. The short testing paradigm based on these results is easy to use in clinical practice.

Machine learning prediction of mental health strategy selection in school aged children using neurocognitive data

This study investigates the use of neurocognitive data collected via functional near infrared spectroscopy (fNIRS) for the development of predictive algorithms related to client Dialectical Behavior Therapy (DBT) skill use in Digital Counseling Environments (DCEs). Specifically, the aim of this study is to examine the utility of using neurocognitive data in combination with a machine learning algorithm to predict client selection of strategies in a virtual environment. Participants were recruited from a rural school located in the United States (n = 50). Student participants engaged in one of three DBT skill development conditions: time-delay-control, face-to-face, and virtual reality-based. Neurocognitive responses in the form of ratios of oxygenated and deoxygenated blood in neural tissue (hemodynamic response) were collected during the learning phase within each condition. The data was then used to train and develop a model to predict the success of client selection of DBT strategies when presented with the opportunity in an assessment phase. The average predictive accuracy of the resultant algorithm was ∼83%. Results also illustrated the potential to capture changes in cognition via hemodynamic response as they occur during DBT skill development in near real-time (∼300 ms). Findings from this study illustrate that the use of neurocognitive data in DCEs may be used to successfully predict client outcomes and increase the quality and reliability of counselor assessments of client progress, aid in the development of artificially intelligent counselors, and improve counselors use of client-based analytics in face-to-face and digital counseling environments.

Integration of genome-wide association and genomic prediction for dissecting seed protein and amino acid in foxtail millet

ContextThe exploitation and utilization of genetic variation in crop germplasm benefit genomics-informed breeding programs. The integration of genome-wide association study (GWAS) and genomic prediction (GP) holds promise for illustrating genetic basis of phenotypic variation. Currently, the use of this integrative approach for dissecting grain protein and amino acid traits remains limited in crop species. ObjectiveThe research objectives were to identify significant single nucleotide polymorphisms (SNPs) associated with seed protein and amino acids and to assess the genomic predictive ability for nutritional traits in natural population of foxtail millet. MethodsIn this study, the concentrations of seed protein and 17 amino acids were assessed in a training population consisting of 238 diverse accessions of foxtail millet. The significant SNPs associated with these phenotypes were identified by GWAS with a mix-linear model. Extensive cross-validation was performed to evaluate the accuracy of genomic prediction of seed nutritional traits. Furthermore, the genomic prediction model was optimized and externally validated in 211 untested materials with the most significant SNPs obtained from GWAS. ResultsA significant genotype variance was detected for seed protein and amino acids and broad-sense heritability (h2) ranged from 0.45 to 0.87 (P < 0.05). The GWAS identified 39 significant SNPs associated with 15 amino acids (p < 1e-06), explaining an average of 12.2% phenotypic variation per locus. The number of significant SNPs ranged from 1 (Asp, Ile, Glu) to 23 for Phe amino acid, including six SNPs associated with one or multiple amino acids. In the training population, the prediction accuracies of protein and amino acids, tested using the 10-fold cross-validation model with whole genome-wide SNPs, ranged from 0.23 (Glu) to 0.60 (Lys) with an average value of 0.44%. Population structure had little effect on the prediction accuracy. In an empirical validation experiment involving 211 untested accessions, the averaged prediction accuracy increased by 12.3% across all traits when using the 1000 top SNPs obtained from GWAS, compared with the same number of randomly selected SNPs from the entire genome. ConclusionsLarge variations of seed protein and amino acids were found in a natural population of foxtail millet. GWAS identified 39 significant SNPs associated with 15 amino acids, and six SNPs were linked to more than one trait. The 10-fold cross-validation model produced the relatively higher prediction accuracies in seed quality traits of diverse accessions. By applying the significant SNPs from GWAS to the untested accessions, the use of 1000 GWAS-top SNPs resulted in more powerful predictions for complex nutritional traits than those using the same number of randomly selected SNPs across the genome. SignificanceThe research highlights the importance and effectiveness of integrative GWAS and genomic prediction in revealing the genetic architecture of seed protein and amino acid variation in foxtail millet. The results provide valuable insights into assessment and improvement of germplasm resources for advancing precision crop phenotyping and enhancing breeding programs for crop nutritional traits.

Multi-objective optimization of machining parameters in complete peripheral milling process with variable curvature workpieces

The mechanism of peripheral milling with variable curvature has yet to be revealed. There is an urgent need to develop a composite optimization approach for the full milling process, including rough and fine machining. The influence of curvature on the energy consumption mechanism in peripheral milling is investigated in this paper. Firstly, the geometric parameters of peripheral milling are analyzed, and the material removal rate calculation method with variable curvature is proposed. The effects of workpiece curvature, machining parameters, and tool wear on energy consumption and surface quality are revealed. Secondly, prediction models for machine tool specific energy and surface roughness are developed in accordance with the examined material removal mechanism. Then, the carbon emission calculation method for the complete peripheral milling process is put forward based on the developed energy consumption mechanism. Finally, a genetic algorithm with elite strategy is used to optimize the multi-dimensional and multi-objective machining parameters. The objective is to achieve the maximum machining efficiency, minimize carbon emissions, and attain excellent surface quality during the complete milling process. The research aims to concurrently optimize machining parameters for rough and fine machining processes, thereby enabling effective and environmentally friendly peripheral milling while maintaining machining quality. The determination coefficient R2 for the proposed model exceeds 0.97, while the average accuracy of predictions surpasses 96 %. The research holds particular importance in providing guidance for peripheral milling operations involving workpieces with variable curvature.

Quality prediction of kiwifruit based on transfer learning

Kiwifruit has a high decay rate, in part because quality changes during storage cannot be easily monitored in real time. In order to better monitor the shelf life of kiwifruit and understand the quality changing process during storage, internal quality indexes such as hardness, respiratory intensity and TSS(Total Soluble Solid) were considered into the prediction models. The prediction models were constructed based on BPNN (Back Propagation Neural Network), Random Forest (RF) and XGBoost (eXtreme Gradient Boosting) respectively. And transfer learning algorithm was used to construct the quality prediction models with BPNN, RF, and XGBoost algorithms as the base learner. In the experiments, sample data were augmented by adding Gaussian noise, which effectively prevented the model from over-fitting. The experimental results showed that the prediction accuracy of each index based on transfer learning was better than that of individual BPNN, RF and XGBoost. Moreover, the average prediction accuracy of the models based on transfer learning was 96.2%, and that of respiratory intensity was as high as 99.4%. Therefore transfer learning can be used to effectively analyze and predict changes of kiwifruit quality indexes during storage.

Harnessing Machine Learning for Predicting Cryptocurrency Returns

The study investigates the predictability of both the individual and basket of 10 major cryptocurrencies’ daily price changes between 2017 and 2023 by employing various machine learning classification algorithms such as random forests, k-nearest neighbour, decision trees, logistic regression, and Bernoulli naïve Bayes. These models utilize 15 different features based on historical price data and technical indicators as input features. The study estimates find logistic regression as superior over other models under consideration in predicting cryptocurrency daily returns. Overall, the study finds that on an average machine learning classification algorithms predictive accuracies have surpassed 50% when applied to daily frequencies on the basket of 10 major cryptocurrencies.

A Comprehensive Evaluation of Shale Oil Reservoir Quality

To enhance the accuracy of the comprehensive evaluation of reservoir quality in shale oil fractured horizontal wells, the Pearson correlation analysis method was employed to study the correlations between geological parameters and their relationship with production. Through principal component analysis, the original factors were linearly combined into principal components with clear and specific physical meanings, aiming to eliminate correlations among factors. Furthermore, Gaussian membership functions were applied to delineate fuzzy levels, and the entropy weight method was used to determine the weights of principal components, establishing a fuzzy comprehensive evaluation model for reservoir quality. Without using principal component analysis, the correlation coefficient between production and evaluation results for the 40 wells in the Cangdong shale oil field was only 0.7609. However, after applying principal component analysis, the correlation coefficient increased to 0.9132. Field application demonstrated that the average prediction accuracy for the cumulative oil production per kilometer of fractured length over 12 months for the 10 applied wells was 91.8%. The proposed comprehensive evaluation method for reservoir quality can guide the assessment of reservoir quality in shale oil horizontal wells.

Short-Term Photovoltaic Power Prediction Based on Extreme Learning Machine with Improved Dung Beetle Optimization Algorithm

Given the inherent volatility and intermittency of photovoltaic power generation, enhancing the precision of photovoltaic power predictions becomes imperative to ensure the stability of power systems and to elevate power quality. This article introduces an intelligent photovoltaic power prediction model based on the Extreme Learning Machine (ELM) with the Adaptive Spiral Dung Beetle Optimization (ASDBO) algorithm. The model aims to accurately predict photovoltaic power generation under multi-factor correlation conditions, including environmental temperature and solar irradiance. The computational efficiency in high-dimensional data feature conditions is enhanced by using the Pearson correlation analysis to determine the state input of the ELM. To address local optimization challenges in traditional Dung Beetle Optimization (DBO) algorithms, a spiral search strategy is implemented during the dung beetle reproduction and foraging stages, expanding the exploration capabilities. Additionally, during the dung beetle theft stage, dynamic adaptive weights update the optimal food competition position, and the levy flight strategy ensures search randomness. By balancing convergence accuracy and search diversity, the proposed algorithm achieves global optimization. Furthermore, eight benchmark functions are chosen for performance testing to validate the effectiveness of the ASDBO algorithm. By optimizing the input weights and implicit thresholds of the ELM through the ASDBO algorithm, a prediction model is established. Short-term prediction experiments for photovoltaic power generation are conducted under different weather conditions. The selected experimental results demonstrate an average prediction accuracy exceeding 93%, highlighting the effectiveness and superiority of the proposed methodology for photovoltaic power prediction.

Deep-learning-based inverse design of colloidal quantum dots

Because optical properties of colloidal quantum dots (QDs) greatly depend on their layer thicknesses at the sub-nanometer scale, accurately predicting optical properties based on these structural parameters can help streamline the development process. However, most of traditional inverse design approaches require time-consuming numerical simulations and optimization processes. In this simulation-only paper, we propose a deep-learning-based inverse design model for colloidal InP/ZnSe/ZnS QDs based on the structure parameters of layer thicknesses and optical properties of the emission and absorption spectrum. Four different deep neural network (DNN) structures are implemented and compared. A tandem DNN structure with the simultaneous use of both emission and absorption spectra as input data shows the best accuracy due to its ability to alleviate the non-uniqueness problem in the inverse design of colloidal QDs. Our inverse design model enables the prediction of layer thicknesses for QDs that exhibit on-demand optical emission/absorption spectra, which are particularly enhanced at the specific target wavelength region. Moreover, the weights and biases of the well-trained tandem DNN for InP/ZnSe/ZnS QDs can be transferred and tuned to implement a transfer-learning-assisted inverse design model for CdSe/ZnSe/ZnS QDs, where the amount of training data to achieve the average prediction accuracy of 99.3 % is reduced by a maximum of 71.4 %.

Risk prediction models for lung cancer in people who have never smoked: a protocol of a systematic review

BackgroundLung cancer is one of the most commonly diagnosed cancers and the leading cause of cancer-related death worldwide. Although smoking is the primary cause of the cancer, lung cancer is also commonly diagnosed in people who have never smoked. Currently, the proportion of people who have never smoked diagnosed with lung cancer is increasing. Despite this alarming trend, this population is ineligible for lung screening. With the increasing proportion of people who have never smoked among lung cancer cases, there is a pressing need to develop prediction models to identify high-risk people who have never smoked and include them in lung cancer screening programs. Thus, our systematic review is intended to provide a comprehensive summary of the evidence on existing risk prediction models for lung cancer in people who have never smoked.MethodsElectronic searches will be conducted in MEDLINE (Ovid), Embase (Ovid), Web of Science Core Collection (Clarivate Analytics), Scopus, and Europe PMC and Open-Access Theses and Dissertations databases. Two reviewers will independently perform title and abstract screening, full-text review, and data extraction using the Covidence review platform. Data extraction will be performed based on the Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies (CHARMS). The risk of bias will be evaluated independently by two reviewers using the Prediction model Risk-of-Bias Assessment Tool (PROBAST) tool. If a sufficient number of studies are identified to have externally validated the same prediction model, we will combine model performance measures to evaluate the model’s average predictive accuracy (e.g., calibration, discrimination) across diverse settings and populations and explore sources of heterogeneity.DiscussionThe results of the review will identify risk prediction models for lung cancer in people who have never smoked. These will be useful for researchers planning to develop novel prediction models, and for clinical practitioners and policy makers seeking guidance for clinical decision-making and the formulation of future lung cancer screening strategies for people who have never smoked.Systematic review registrationThis protocol has been registered in PROSPERO under the registration number CRD42023483824.

Detection and mapping of Amaranthus spinosus L. in bermudagrass pastures using drone imagery and deep learning for a site‐specific weed management

AbstractWeed encroachment negatively affects pasture productivity by reducing herbage allowance, stocking rates, and livestock performance. Amaranthus spinosus L. is a weed species widely found in pastures worldwide and is considered challenging for ranchers due to its great potential for invasion, making it difficult to control. The high costs of chemical application and the global concern about environmental impacts restrict indiscriminate herbicide spraying in pastures. Site‐specific weed management (SSWM) is a weed management strategy based on weed spot‐spraying that has the potential to overcome these issues. Images from unmanned aerial vehicles (UAVs) can provide valuable information for weed mapping to drive the herbicide application in pastures. Deep learning techniques have been highlighted in image classification tasks. We developed a deep convolutional neural network (CNN)‐based image segmentation model based on the U‐Net architecture to detect and map Amaranthus spinosus in bermudagrass pastures using red–green–blue images acquired through UAV flying in moderate‐high altitude. The images were acquired from twelve paddocks under three treatments (weed‐free, weed‐strips, or weed‐infested) during the summer (2021‐2022). The CNN model was able to detect around 80% of the A. spinosus with an average prediction accuracy of 94%. Our weed mapping showed the potential of using the U‐Net model to generate a herbicide application map to be inserted into the sprayer system, reducing up to 76% of the amount of herbicide applied. Further studies are encouraged to increase the robustness of the model across species and development stages and develop sprayer systems to implement the spot‐spraying in field conditions.

Unified mRNA Subcellular Localization Predictor based on machine learning techniques

BackgroundThe mRNA subcellular localization bears substantial impact in the regulation of gene expression, cellular migration, and adaptation. However, the methods employed for experimental determination of this localization are arduous, time-intensive, and come with a high cost.MethodsIn this research article, we tackle the essential challenge of predicting the subcellular location of messenger RNAs (mRNAs) through Unified mRNA Subcellular Localization Predictor (UMSLP), a machine learning (ML) based approach. We embrace an in silico strategy that incorporate four distinct feature sets: kmer, pseudo k-tuple nucleotide composition, nucleotide physicochemical attributes, and the 3D sequence depiction achieved via Z-curve transformation for predicting subcellular localization in benchmark dataset across five distinct subcellular locales, encompassing nucleus, cytoplasm, extracellular region (ExR), mitochondria, and endoplasmic reticulum (ER).ResultsThe proposed ML model UMSLP attains cutting-edge outcomes in predicting mRNA subcellular localization. On independent testing dataset, UMSLP ahcieved over 87% precision, 94% specificity, and 94% accuracy. Compared to other existing tools, UMSLP outperformed mRNALocator, mRNALoc, and SubLocEP by 11%, 21%, and 32%, respectively on average prediction accuracy for all five locales. SHapley Additive exPlanations analysis highlights the dominance of k-mer features in predicting cytoplasm, nucleus, ER, and ExR localizations, while Z-curve based features play pivotal roles in mitochondria subcellular localization detection.AvailabilityWe have shared datasets, code, Docker API for users in GitHub at: https://github.com/smusleh/UMSLP.

Research on predicting football matches based on handicap data and BPNN

Football is one of the most influential sports in the world, and billions of people around the globe pay much attention to the football matches. With the growing popularity of football and the continuous development of the football betting industry, the prediction of the outcomes of football matches has become a hot topic in the commercial operations of sports especially footballs in recent years. It is also an important subject of academic research. In this paper, we develop a football match result prediction model based on the back propagation neural network. We take the German Bundesliga competitions as the research object in this paper. In addition to utilizing historical statistic data and team attributes from previous matches, we also incorporate a new dataset, known as handicap data, which refers to the odds data of the football matches, as the input layer of the BPNN (back propagation neural networks) for prediction. We also innovatively use varying numbers of hidden nodes, which greatly improves the prediction accuracy and stability of the model. Experimental results indicate that the average prediction accuracy of this football match prediction model is around 57.2%, with the highest prediction accuracy reaching 59.8% and the lowest prediction accuracy at 53.8%. The prediction model demonstrates relative stability, with no significant fluctuations in prediction accuracy.

Large-scale comparison of machine learning methods for profiling prediction of kinase inhibitors

Conventional machine learning (ML) and deep learning (DL) play a key role in the selectivity prediction of kinase inhibitors. A number of models based on available datasets can be used to predict the kinase profile of compounds, but there is still controversy about the advantages and disadvantages of ML and DL for such tasks. In this study, we constructed a comprehensive benchmark dataset of kinase inhibitors, involving in 141,086 unique compounds and 216,823 well-defined bioassay data points for 354 kinases. We then systematically compared the performance of 12 ML and DL methods on the kinase profiling prediction task. Extensive experimental results reveal that (1) Descriptor-based ML models generally slightly outperform fingerprint-based ML models in terms of predictive performance. RF as an ensemble learning approach displays the overall best predictive performance. (2) Single-task graph-based DL models are generally inferior to conventional descriptor- and fingerprint-based ML models, however, the corresponding multi-task models generally improves the average accuracy of kinase profile prediction. For example, the multi-task FP-GNN model outperforms the conventional descriptor- and fingerprint-based ML models with an average AUC of 0.807. (3) Fusion models based on voting and stacking methods can further improve the performance of the kinase profiling prediction task, specifically, RF::AtomPairs + FP2 + RDKitDes fusion model performs best with the highest average AUC value of 0.825 on the test sets. These findings provide useful information for guiding choices of the ML and DL methods for the kinase profiling prediction tasks. Finally, an online platform called KIPP (https://kipp.idruglab.cn) and python software are developed based on the best models to support the kinase profiling prediction, as well as various kinase inhibitor identification tasks including virtual screening, compound repositioning and target fishing.

Identification of carbonate sedimentary facies from well logs with machine learning

Sedimentary facies identification is critical for carbonate oil and gas reservoir development. The traditional method of sedimentary facies identification not only be affected by the engineer's experience but also takes a long time. Identifying carbonate sedimentary facies based on machine learning is the trend of future development and has the advantages of short time consuming and reliable results without engineers' subjective influence. Although many references reported the application of machine learning to identify lithofacies, but identifying sedimentary facies of carbonate reservoirs is much more challenging due to the complex sedimentary environment and tectonic movement. This paper compares the performance of the carbonate sedimentary facies identification using four different machine learning models, and the optimal machine learning with the highest prediction accuracy is recommended. First, the carbonate sedimentary facies are classified into the lagoon, shallow sea, shoal, fore-shoal, and inter-shoal five tags based on the well loggings. Then, five well log curves including spectral gamma ray (SGR), uranium-free gamma ray (CGR), photoelectric absorption cross-section index (PE), true formation resistivity (RT), shallow lateral resistivity (RS) are used as the input, and the manual identified carbonate sedimentary facies are used as the output of the machine learning model. The performance of four different machine learning algorithms, including support vector machine (SVM), deep neural network (DNN), long short-term memory (LSTM) network, and random forest (RF) are compared. The other two wells are used for model validation. The research results show that the RF method has the highest accuracy of sedimentary facies prediction, and the average prediction accuracy is 78.81%; the average accuracy of sedimentary facies prediction using SVM is 77.93%. The sedimentary facies predictions using DNN and LSTM are less satisfying compared with RF and SVM, and the average accuracy is 69.94% and 73.05%, respectively. The predicted carbonate sedimentary facies by LSTM are more continuous compared with other machine learning models. This study is helpful for identifying compelx sedimentary facies of carbonate reservoirs from well logs.

Revealing the failure mechanism of high strength seawater coral aggregate reinforced concrete slabs based on acoustic emission technology: Parameter analysis and DBN-BPNN classification

In this study, high strength seawater coral aggregate concrete (SCAC) with a compressive strength of 80 MPa was prepared, and seawater coral aggregate reinforced concrete slab (SCARCS) was designed with epoxy-coated steel bars. Combining with the AE technology, monotonic flexural static tests were conducted on SCARCS specimens with three varying reinforcement ratios (i.e., 0.85%, 1.13%, and 1.41%). Results showed that the increasing the reinforcement ratio led to an increase in peak bearing capacity and a decrease in center deflections at the peak. The loading process can be divided into four stages based on the varying magnitudes of the b-value and Ib-value. The primary mode of failure was tensile microcracks at various stages, and the ratio of shear microcracks reached its maximum value in the final stage of loading. Besides, based on the signal intensity analysis, it can be observed that the scale and complexity of cracks propagation inside the SCARCS specimen increased as the reinforcement ratio rose. In addition, back propagation neural network (BPNN) classification model optimized by deep belief network (DBN) algorithm was proposed to predict the damage degree of SCARCS during loading, and the average prediction accuracy of the proposed model was shown to be over 89%. The classification outcomes demonstrated that with an increase in the reinforcement ratio of SCARCS, the proportion of AE signals in Cluster 3 increased, while the proportion of AE signals in Cluster 2 dropped. Notably, both the average frequency and energy presented an ascending trend. The findings of this study provided valuable insights for designing and assessing the safety of SCAC structures for essential island and reef engineering infrastructure.

A review of icing prediction techniques for four typical surfaces in low-temperature natural environments

As a common phenomenon in nature and industrial fields, icing always adversely affects life, and plays negative effects. Although series of anti-/de-icing techniques are widely investigated, icing prediction techniques on surfaces under low-temperature natural environment are more important due to effectively reduce or even prevent icing caused harms. To understand the current research progress of icing prediction techniques, four typical static and moving surfaces that are prone to ice formation were selected as the research objects in this study, including road, transmission line, wind turbine blade and aircraft surfaces. As summarized, prediction methods mainly include physical models, statistical models, and machine learning. For road surface, statistical analysis, theoretical analysis and data mining methods are widely used, with the average prediction accuracy reaching 80%. For wind turbine blade and transmission line surfaces, both model-driven and data-driven methods were used, resulting in average prediction accuracies of 80% and 90%, respectively. For aircraft surface, the method of numerical analysis combined with machine learning is the mainstay, with a deviation of less than 20%. Summary and outlook are finally given, which are beneficial for the optimization of ice prediction technology in different industrial fields.