Input Dataset Research Articles

Cucconi Feature Extracted Random Decision Forest Classification for Efficient Sentiment Analysis

Sentiment is the method of conveying the opinion by the customer for particular product e-commerce website. Sentiment analysis is otherwise termed as opinion mining that identifies the polarity of extracted opinions. Nowadays, opinion mining concept received large interest by many researchers to identify the polarity of the statements. A number of scholars have conducted their research on efficient sentiment analysis with better accuracy. But the space complexity and time complexity performance did not reduced using existing classification techniques. To deal with these issues, CFDFC (Cucconi Feature Extracted Random Decision Forest Classification) Approach is presented. The primary aim of CFDFC Approach is to perform effective sentiment analysis with improved accuracy and reduced time complexity. The number of user review statement is collected from the input dataset. The proposed CFDFC Technique comprises of pre-processes, feature extractions, and classifications. Pre-processes eliminate stop words and stem words from user reviews. After pre-processing, the features (i.e. keywords) extraction process is performed to minimize the dimensionality and time consumption for opinion classification. This work uses cucconi projective feature extractions in CFDFC to squeeze out opinion words from reviews where cucconi projective feature extraction is generally a dimensionality reduction technique used for identifying the projections in multidimensional data. Finally, the classification process is carried out using random decision forest classifier. The random decision forest classifier uses the ID3 DT (decision tree) as a weak learner to classify the review statements with the extracted features. By using the voting scheme, random decision forest classifications finally classify instances with higher accuracy. This again is useful to boost the performance of sentiment analysis. The comprehensive experimental assessment is carried out using various parameters including accuracy, error rates, recall values , time and space complexities with regard to the number of review statements gathered from the dataset.The statistical results obtained show that the proposed CFDFC Technique achieves remarkable accuracy, recall, and minimal time complexity than the current methods.

Mathematical Model of an Automated System for Learning to Play the Piano

Purpose of research. Nowadays, the automation of the distance learning process or self-study is relevant. However, to increase its effectiveness, there is a need for a personality-oriented approach to the student. The creation of adaptive learning technologies is becoming a priority. Development of a mathematical model of an automated system for learning to play the piano was the purpose of this work.Methods. The article discusses one of the approaches to modeling educational processes and systems - the cybernetic approach. The use of this approach allows us to consider the model of the educational system from the point of view of internal blocks and their functions, it allows us to identify sets of input and output data and internal relationships of various components of the system. The mathematical model is based on the operations of the algebra of logic and set theory.Resalts. Based on the cybernetic approach, the article provides a functional diagram of the system, which clearly demonstrates the sequence of actions performed, as well as the flow of various sets of input and output data. The functionality of the blocks is described in the terminology of operators, which determine the sequence and the required number of questionnaire questions to determine the initial knowledge, skills, and abilities of the trainee and his goals, select the most optimal initial training plan, assess progress and adjust the initial training plan depending on the results of the student. For each operator, sets of inputs and outputs are defined. The description of the data shows what information the system must operate in order for it to work correctly and efficiently. Another significant result of the work was the mathematical model of the system obtained as a result of research, which determines the relationship between input and output data.Conclusion. The cybernetic approach to modeling educational activities, including in the system of additional education, allows you to structure complex logical connections in the educational system, to determine the sequence and options for the system. The model will form the basis of the information system for piano playing teaching.

Automated and intelligent system for World Health Organization data forecasting

AbstractTechnological advances and social transformations have enabled the circulation of a large amount of data in the health area. Analyzing this data becomes more critical and more challenging as the volume of data increases. An alternative to performing this analysis is to use data analysis techniques to process input data sets and build consistent databases for input to machine learning algorithms. Thus, it can forecast future scenarios and collaborate with knowledge discovery. In this context, this work aims to develop a parameterizable system with automated decisions capable of collecting and analyzing many indicators provided by the World Health Organization (WHO). After these analysis steps, the system applies machine learning algorithms for predictions of different indicators to process information automatically, finding different knowledge discovery scenarios. Thus, the contribution of this article is an automated and intelligent system for WHO data forecasting. The efficiency of the system's choices and forecasts was proven with experiments in five different areas of health, obtaining assertiveness by up to 99.92%, root‐mean‐square error (RMSE) by up to 0.0286, and Kling‐Gupta Efficiency (KGE) by up to 0.9988, hitting even the most complex cases, as shown in the confusion matrices. Finally, three case studies were carried out to expand the studies and present the potential of the system in different contexts: anemia in children, age‐standardized suicide rates in men, and number of road traffic deaths.

Estimation of porosity and volume of shale using artificial intelligence, case study of Kashafrud Gas Reservoir, NE Iran

The key problem in oil exploration and engineering is the lack of accurate and reliable data about the reservoir parameters of a field. Having a precise assessment of petrophysical properties can provide the ability to make decisions with a high degree of confidence about planning for production, exploitation, and further field development scenario. In this research, an artificial intelligence (AI)-based approach was developed to improve the estimation of reservoir parameters including porosity and volume of shale, which has a significant role in different stages of hydrocarbon exploration, in the Kashafrud Gas Reservoir in the northeast of Iran. For this purpose, we measured the petrophysical properties of 27 samples of the Kashafrud Formation. To increase the amount of data for employing a multilayer perceptron (MLP) artificial neural network (ANN), a geostatistical algorithm was used to increase the amount of laboratory measured data of porosity and volume of shale to 686 and 702, respectively. In addition, 2263 well-logging data from the same well were provided. The optimal MLP network with the topology of 6-7-1, and 6-8-1 was selected to estimate the porosity and shale volume with mean squared error (MSE) of 2.78731E−4, and 1.28701E−9, respectively. The training process was performed using two different sets of input data. In the first approach, all available well-logging data were used as input, ending up in high MSE. In the second approach, some selected well logs were used based on the results of sensitivity analysis which clearly improved the estimations. The ability of MLP networks made great improvements in the estimation of the both parameters up to 99.9%. The presence of valuable core data in this study significantly improved the process of comparison and conclusion. The final results prove that AI is a trusted method, also the potential of the ANN method for the reservoir characterization and evaluation associated problems should be taken into consideration. Due to the unavailability of core data along the whole wells, the application of intelligent methods, such as machine learning (ML) can be used to estimate the parameters in other oil or gas fields and wells.

Development of mathematical models to support decision-making regarding the functioning of critical infrastructure in the industry of energy supply

The object of research is the energy supply company and the processes of generation and supply of electric energy. The paper examines the problems of building mathematical models for forecasting the operation of a critical infrastructure object in the conditions of a changing security environment, characterized by unpredictability, the presence of uncertainties of various types, the appearance of new threats, their combinations, changes in the form, duration, nature of their influence. In the work, the main attention is paid to the study of the functioning of critical infrastructure in the field of energy supply. Based on the study of the functioning of the energy company system, methods of dealing with uncertainties at the stage of data preparation for analysis and during the preliminary construction of models are presented, in particular, statistical and probabilistic approaches, modeling of the studied processes, alternative methods of estimating model parameters, etc. The complexity of preparing the input data set is related to the fact that it is necessary to ensure the representativeness and variability of the data sets, given that a significant number of factors must be included in the model according to the requirements of regulatory documents, which can lead to multicollinearity of the input variables. The paper proposes an analytical toolkit based on the use of mathematical models and their combinations, intended for use in specialized decision support systems. In the course of the research, a number of numerical experiments were conducted, in which the proposed methodology was worked out on the materials of the enterprise – the object of the critical infrastructure of the energy sector. SAS Energy Forecasting software was used to build the models. The best forecasting results are obtained using generalized linear models (GLM), in particular the GLM model in the form of ARIMAX (a moving average autoregressive model that includes an integrated trend component and external regressors). The proposals presented in the work will allow to increase the efficiency of the functioning of the energy sector, including the determination of the goals, tasks and benchmarks of its operation in regular mode for certain periods of time, as well as in the field of development of universal and special mechanisms for ensuring stability in the mode of response to threats and critical situations.

BURNT AREAS SEMANTIC SEGMENTATION FROM SENTINEL DATA USING THE U-NET NETWORK TRAINED WITH SEMI-AUTOMATED ANNOTATIONS

Abstract. The Pantanal biome is one of the most important wetlands on the planet, harboring a rich biodiversity whilst being critical in maintaining hydrological cycles and climate regulation. However, the occurrence of fires in the biome has represented a significant threat to this unique ecosystem and its multiple functions. Understanding the extent, intensity and environmental impacts caused by fires in the Pantanal, is of unique importance for the preservation of the biome's biodiversity. Remote sensing techniques have played an important role in detecting and mapping burnt areas, especially SAR (Synthetic Aperture Radar) orbital systems, that are able to collect data in regions with frequent cloud cover or during extreme fire events. In this context, the objective of this study was to evaluate the potential of the U-Net semantic segmentation network applied to SAR data in the detection of burnt areas in the Brazilian Pantanal. For this, a semi-automatic annotated dataset was generated and considered as ground truth to evaluate the result obtained by the network. Two input datasets were evaluated in the detection of burnt areas, one containing optical and SAR data whereas the other containing only SAR data. The predictions of the two datasets were consistent with the semi-automatically generated annotation, showing similar spatial distribution but presenting a greater number of burnt areas. The model using both optical and SAR data achieved IoU (Intersection of Union) of 0.69 whereas the SAR only model had 0.60. Considering the amount of available data and the complexity of burnt area detection, the predictions achieved were adequate.

Development of a source-specific biomass burning emissions inventory for Washington, Oregon, and California

While wildfires are the dominant source of biomass burning emissions across the U.S., prescribed and agricultural burns also contribute to emissions and ambient air quality. Understanding the differential impacts that these different sources of biomass burning have on air quality and public health is important in the development of effective exposure reduction strategies. We leverage the high-resolution Fire INventory from NCAR (FINN) to develop a source-specific biomass burning emission inventory that distinguishes between wildfire, prescribed burn, and agricultural burn sources across Washington, Oregon, and California. We describe the methodology for reclassifying FINN using multiple national and state-level fire and fuel treatment inventories and present the results of the emissions reclassification process for these three states from 2012 to 2020. We matched 72%, 78% and 81% of FINN emissions event locations to fire locations from the fire type input datasets in Washington, Oregon, and California, respectively. When compared to county-level fire type-specific PM2.5 emissions estimates from the 2014, 2017, and 2020 EPA National Emissions Inventory, we found greater agreement among wildfire-specific estimates across years and states (R2 range: 0.59–0.98), relative to prescribed (R2 range: 0.12–0.73) and agricultural burns (R2 range: 0.43–0.95). This fire type-specific biomass burning emissions inventory can be used in air pollution transport models to better understand the impacts of each of these sources on air quality across the region along with downstream public health impacts.

Identifying Crop Growth Stages from Solar-Induced Chlorophyll Fluorescence Data in Maize and Winter Wheat from Ground and Satellite Measurements

Crop growth stages are integral components of plant phenology and are of significant ecological and agricultural importance. While the use of remote sensing methods for phenology identification in cropland ecosystems has been extensively explored in previous studies, the focus has often been on land surface phenology, primarily related to the start and end of the growing season. In contrast, the monitoring of crop growth within an agronomic framework has been limited, particularly in the context of recently developed solar-induced chlorophyll fluorescence (SIF) data. Additionally, some critical growth stages have not received adequate attention or evaluation. This study aims to assess the utility of SIF data, collected from both ground and satellite measurements, for identifying critical crop growth stages within the realm of remote sensing phenological estimation. A comparative analysis was conducted using enhanced vegetation index (EVI) data at the Shangqiu site in the North China Plain from 2018 to 2022. Both SIF and EVI time-series data, obtained from ground and satellite sources, undergo a comprehensive phenological estimation framework encompassing pre-processing, modeling, and transition characterization. This approach involves reconciling time-series phenological patterns with crop growth stages, revealing the necessity of redefining the mapping relationship between these two fundamental concepts. After preprocessing the time-series data, the framework incorporates the phenological modeling process employing two double logistic models and a spline model for comparison. Additionally, it includes phenological transition characterization using four different methods. Consequently, each input dataset undergoes an assessment, resulting in 12 sets of estimations, which are compared to select the ideal estimation portfolio for identifying the growth stages of maize and winter wheat. Our findings highlight the efficacy of SIF data in accurately identifying the growth stages of maize and winter wheat, achieving remarkable results with an R-square exceeding 0.9 and an RMSE of less than 1 week for key growth stages (KGSs). Notably, SIF data demonstrate superior accuracy, robustness, and sensitivity to phenological events when compared to EVI data. This study establishes an estimation portfolio utilizing SIF data, involving the Gu model, a double logistic model, as the preferred phenological modelling method together with various compositing methods and transition characterization methods, suitable for most KGSs. These findings create opportunities for future research aimed at enhancing and standardizing crop growth stage identification using remote sensing data for a wide range of KGSs.

Evaluation of gridded precipitation data in water availability modeling in CONUS

Water availability estimates are affected by the quality of data used. In this study precipitation data from meteorological forcing datasets specifically compiled for large-scale hydrological modeling are evaluated over the Conterminous United States (CONUS). The performance of the datasets was evaluated using Pearson Correlation (r), Nash Sutcliff Efficiency (NSE), Percent Bias (PBIAS), and Kling Gupta Efficiency (KGE) criterion. Three observation-based gridded precipitation data from the Inter-Sectoral Impact Model Intercomparison Project – ISIMIP (EartH2Observe, WFDEI and ERA-Interim data Merged and Bias-corrected for ISIMIP (EWEMBI), WATCH Forcing Data methodologyapplied to ERA-Interim data (WFDEI), and WFDE5 over land merged with ERA5 over the ocean (W5E5)) and one regional dataset (GRIDMET) were compared. The datasets were compared using more than 3300 gage precipitation observations. For grid-level comparisons, the Climate Prediction Center's Unified gage-based gridded precipitation dataset (CPC) was used as a reference dataset after verification with gage observations. The grid-to-grid comparison was made on daily, monthly, annual, and seasonal time steps. Grid runoff and discharge were simulated using the Community Water Model (CWatM) model. The flow estimations using the datasets were compared with CPC estimations and with more than 900 discharge observations distributed over the nine climate regions of CONUS. Furthermore, the per capita water availability was estimated and compared. In comparison with gage precipitation observation, CPC had superior performance, followed by GRIDMET. CPC resulted in daily KGE, NSE, and r above 0.5, and PBIAS within 25 % in 79 %, 69 %, 87 %, and 90 % of the stations, respectively. Grid-to-grid precipitation comparisons resulted in overall comparable performance between the datasets. In grid discharge estimation, GRIDMET (median KGE = 0.68) had the highest correlation with CPC. WFDEI, W5E5, and EWEMBI show close performance to each other in estimating grid discharge with median KGE values of 0.43, 0.25, and 0.45 respectively. The performance of the datasets as compared to observed stream flow gages also showed comparable performance with above zero KGE in 73, 56, 71, 60, and 70 stations by GRIDMET, EWEMBI, W5E5, and WFDEI, and CPC respectively. In estimating water availability per capita, W5E5 showed an overestimation as compared to ISMIP datasets. The estimations from CPC and GRIDMET were close to each other. This study serves as a valuable resource for guiding the selection of a suitable dataset for large-scale modeling over the CONUS. By providing performance information for the compared datasets, it assists in making informed data selection and contributes to our understanding of how the selection of input datasets affect water availability modeling.

Hough Transform based Deep Belief Network and Improved Homomorphic Encryption for Cloud Security based Intrusion Discovery

The enlarge development in information technology is cloud computing, which offers minimized infrastructure cost, lower maintenance, greater flexibility and scalability. Nowadays, the network security plays vital role in enterprises and organizations. The influence vulnerabilities were occurred due to attackers based on network configuration. Because of cloud and IoT growth, enlarge amount of data obtained from IoT sensor and devices are transmitted to cloud data centers. Several security issues like focused web servers in the cloud and information collection mishandling are faced by storage and cloud-based computing when offering us considerable convenience. For that reason, this article proposes a deep learning-based cloud security oriented intrusion discovery. Primarily, the input dataset is pre-processed by using normalization techniques followed by the features are selected using an Adaptive White Shark Optimization (AWSO) algorithm. The normal and intrusion data is classified by using Hough Transform based Deep Belief Network (HT-DBN) after that the sensitive data are secured with the help of an Improved Homomorphic Encryption (IHE) model. The simulation tool of MATLAB is been used to simulate the proposed implementation part and the experimental results outperformed the detection accuracy of 97% than other previous approaches.

Hough Transform based Deep Belief Network and Improved Homomorphic Encryption for Cloud Security based Intrusion Discovery

The enlarge development in information technology is cloud computing, which offers minimized infrastructure cost, lower maintenance, greater flexibility and scalability. Nowadays, the network security plays vital role in enterprises and organizations. The influence vulnerabilities were occurred due to attackers based on network configuration. Because of cloud and IoT growth, enlarge amount of data obtained from IoT sensor and devices are transmitted to cloud data centers. Several security issues like focused web servers in the cloud and information collection mishandling are faced by storage and cloud-based computing when offering us considerable convenience. For that reason, this article proposes a deep learning-based cloud security oriented intrusion discovery. Primarily, the input dataset is pre-processed by using normalization techniques followed by the features are selected using an Adaptive White Shark Optimization (AWSO) algorithm. The normal and intrusion data is classified by using Hough Transform based Deep Belief Network (HT-DBN) after that the sensitive data are secured with the help of an Improved Homomorphic Encryption (IHE) model. The simulation tool of MATLAB is been used to simulate the proposed implementation part and the experimental results outperformed the detection accuracy of 97% than other previous approaches.

Strong aerosol cooling alone does not explain cold-biased mid-century temperatures in CMIP6 models

Abstract. The current generation of global climate models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) exhibits a surprisingly cold-biased ensemble-mean mid-20th century global-mean surface temperature anomaly, compared to the previous generation Phase 5 (CMIP5) and to the observed mid-century (1940–1970) temperature anomaly. Most CMIP6 models, 31 of 36 models in contrast to 17 of 27 CMIP5 models, are colder than the uncertainty range of the observed anomaly, indicating that the CMIP6 suppressed warming is not caused by a few cold models. However, no clear cause that sufficiently explains the tendency towards suppressed mid-20th century warming emerges. Whereas models that best match observations exclusively exhibit weaker aerosol forcing than that exhibited by colder models, there is not a clear relationship between mid-century temperatures and aerosol forcing. Likewise, no systematic differences emerge among other model aerosol representations, such as inclusion of aerosol–cloud interactions for ice clouds in the model or the type of aerosol model input data set used, nor variations in greenhouse gas forcing or climate sensitivity, that could explain the suppressed warming. This indicates the presence of another cause, or more likely a set of causes, of the suppressed warming in many CMIP6 models. Thus, the prospect of a strong constraint on present-day aerosol forcing based on the mid-century warming is weakened, even if it is encouraging that those models that do match the observed warming best all have relatively weak aerosol forcing.

Method of information technology for structure analysis of urban network fire-rescue units

The subject of this study is the process of analyzing the structure of the network of fire-rescue units of the city in the context of optimizing their spatial distribution. The purpose of this work is to increase the objectivity of decisions made while forming a city network of fire-rescue units by creating an information technology (IT) method for analysis of its structure based on the use of spatially distributed data. Objectives: to find ways to improve the level of fire safety, analyze existing approaches to the formation of a network of fire-rescue units, considering the peculiarities of building and organizing a network, and adapt the classical problem of placement; to propose a method for solving it to minimize the distance between the fire-rescue unit and the possible place of fire while ensuring maximum coverage of the territory by fire service; and to develop IT structure for its implementation based on the information flow model of the process of analyzing the fire-rescue units network using a geospatial approach. The following results were obtained. The study of the classical location problem and its adaptation to real problems arising from the analysis of the urban network of fire-rescue units made it possible to represent it as a set of independent complete bipartite graphs. To search the location of network nodes while solving an adapted problem, an IT method is developed, which, based on the p-median model, combines the author’s methodology for studying information processes and methods of geospatial analysis. Summarizing the requirements of the current legislation, a set of input and output IT data and a set of operations have been formed. The representation of the IT structure in the form of a data flow diagram explains how the set of factors is processed and generalized when making decisions on the creation and / or improvement of the existing city fire department. Conclusions. The results of the bibliographic search confirm the need to consider the spatial features of the area where it is planned to create a fire-rescue unit, as well as the spatial configuration of the urban network of existing fire stations, to evaluate its effectiveness using an integrated indicator. This requires the development of specialized methods focused on the use of geo-information systems for their implementation in decision support systems. Scientific and methodological support for IT has been developed, which gives local authorities a tool for analyzing fire safety in the city to create and / or improve the existing fire protection. An experiment to study the capabilities of the proposed method based on volunteered geographic information on Kharkiv city showed the effectiveness of its use for solving classical problems of placement, considering the accepted restrictions on the spatial availability of fire-rescue units. At the same time, additional opportunities appear in the formation of options for improving the network of fire-rescue units, considering their spatial distribution, workload, accessibility, and the resulting areas of coverage / non-coverage by the fire service. For example, the fire service coverage area of the existing structure of fire stations has been assessed. During the regulated time, it reaches 70% of the Kharkiv city area, and depending on the real road traffic, it can vary from 64.61% to 73.44%. It is illustrated that the creation of two additional fire and rescue units in the northern and southern parts of Kharkiv will increase the coverage area by approximately 5%, on average increasing it to 75.1%

Performance analysis and ANFIS computing of an unreliable Markovian feedback queueing model under a hybrid vacation policy

This study covers the stationary analysis of an unreliable Markovian queueing system with feedback based on a hybrid vacation policy. Customers arrive at the station at random using a Poisson process. Service time is distributed exponentially. The necessary and sufficient requirements for system stability have been demonstrated. Many performance metrics of this queueing system, such as the stationary queue length distribution, have been obtained from the steady-state probability distribution of the queueing model and the steady-state queue length using the matrix geometric method. To establish the model’s performance indicators, a few formulas have also been developed. The Adaptive Neuro-Fuzzy Inference System (ANFIS) was used to develop rules from the input datasets and generate different performance indices. Finally, the impact of system parameters and cost analysis was studied using numerical examples.

Image factory: A method for synthesizing novel CT images with anatomical guidance.

Deep learning in medical applications is limited due to the low availability of large labeled, annotated, or segmented training datasets. With the insufficient data available for model training comes the inability of these networks to learn the fine nuances of the space of possible images in a given medical domain, leading to the possible suppression of important diagnostic features hence making these deep learning systems suboptimal in their performance and vulnerable to adversarialattacks. We formulate a framework to address this lack of labeled data problem. We test this formulation in computed tomographic images domain and present an approach that can synthesize large sets of novel CT images at high resolution across the full Hounsfield (HU)range. Our method only requires a small annotated dataset of lung CT from 30 patients (available online at the TCIA) and a large nonannotated dataset with high resolution CT images from 14k patients (received from NIH, not publicly available). It then converts the small annotated dataset into a large annotated dataset, using a sequence of steps including texture learning via StyleGAN, label learning via U-Net and semi-supervised learning via CycleGAN/Pixel-to-Pixel (P2P) architectures. The large annotated dataset so generated can then be used for the training of deep learning networks for medical applications. It can also be put to use for the synthesis of CT images with varied anatomies that were nonexistent within either of the input datasets, enriching the dataset evenfurther. We demonstrate our framework via lung CT-Scan synthesis along with their novel generated annotations and compared it with other state of the art generative models that only produce images without annotations. We evaluate our framework effectiveness via a visual turing test with help of a few doctors and radiologists. We gain the capability of generating an unlimited amount of annotated CT images. Our approach works for all HU windows with minimal depreciation in anatomical plausibility and hence could be used as a general purpose framework for annotated data augmentation for deep learning applications in medical imaging.

State-of-the-Art Features for Early-Stage Detection of Diabetic Foot Ulcers Based on Thermograms.

Diabetic foot ulcers represent the most frequently recognized and highest risk factor among patients affected by diabetes mellitus. The associated recurrent rate is high, and amputation of the foot or lower limb is often required due to infection. Analysis of infrared thermograms covering the entire plantar aspect of both feet is considered an emerging area of research focused on identifying at an early stage the underlying conditions that sustain skin and tissue damage prior to the onset of superficial wounds. The identification of foot disorders at an early stage using thermography requires establishing a subset of relevant features to reduce decision variability and data misinterpretation and provide a better overall cost-performance for classification. The lack of standardization among thermograms as well as the unbalanced datasets towards diabetic cases hinder the establishment of this suitable subset of features. To date, most studies published are mainly based on the exploitation of the publicly available INAOE dataset, which is composed of thermogram images of healthy and diabetic subjects. However, a recently released dataset, STANDUP, provided data for extending the current state of the art. In this work, an extended and more generalized dataset was employed. A comparison was performed between the more relevant and robust features, previously extracted from the INAOE dataset, with the features extracted from the extended dataset. These features were obtained through state-of-the-art methodologies, including two classical approaches, lasso and random forest, and two variational deep learning-based methods. The extracted features were used as an input to a support vector machine classifier to distinguish between diabetic and healthy subjects. The performance metrics employed confirmed the effectiveness of both the methodology and the state-of-the-art features subsequently extracted. Most importantly, their performance was also demonstrated when considering the generalization achieved through the integration of input datasets. Notably, features associated with the MCA and LPA angiosomes seemed the most relevant.

Adversarial attack driven data augmentation for medical images

An important stage in medical image analysis is segmentation, which aids in focusing on the required area of an image and speeds up findings. Fortunately, deep learning models have taken over with their high-performing capabilities, making this process simpler. The deep learning model’s reliance on vast data, however, makes it difficult to utilize for medical image analysis due to the scarcity of data samples. Too far, a number of data augmentations techniques have been employed to address the issue of data unavailability. Here, we present a novel method of augmentation that enabled the UNet model to segment the input dataset with about 90% accuracy in just 30 epochs. We describe the us- age of fast gradient sign method (FGSM) as an augmentation tool for adversarial machine learning attack methods. Besides, we have developed the method of Inverse FGSM, which im- proves performance by operating in the opposite way from FGSM adversarial attacks. In comparison to the conventional FGSM methodology, our strategy boosted performance up to 6% to 7% on average. The model became more resilient to hostile attacks because to these two strategies. An innovative implementation of adversarial machine learning and resilience augmentation is revealed by the overall analysis of this study.

Parameter estimation in the pollutant dispersion problem with Physics-Informed Neural Networks

In this work, the inverse problem of parameter estimation in the advection-dispersion-reaction equation, modelling the pollutant dispersion in a river, is studied with a Neural Network approach. In the direct problem, the dispersion, velocity and reaction parameters are known and then the initial and boundary value problem is solved by classical numerical methods, where it is used as input dataset for the inverse problem and formulation. In the inverse problem, we know the dispersion and the velocity parameters and also the information about the pollutant concentration from the synthetic experimental data, and then the aim is to estimate the reaction parameter in the advection-dispersion-reaction equation. This inverse problem is solved by an usual Artificial Neural Network (ANN) and by a Physics-Informed Neural Network (PINN), which is a special type of neural networks that includes in its formulation the physical laws that describe the phenomena involved. Numerical experiments with both the ANN and PINN are presented, demonstrating the feasibility of the approach considered.

Surface sediment classification using a deep learning model and unmanned aerial vehicle data of tidal flats

This study proposes a deep learning model, U-Net, to improve surface sediment classification using high-resolution unmanned aerial vehicle (UAV) images. We constructed training datasets with UAV images and corresponding labeling data acquired from three field surveys on the Hwangdo tidal flat. The labeling data indicated the distribution of surface sediment types. We compared the performance of the U-Net model trained in various implementation environments, such as surface sediment criteria, input datasets, and classification models. The U-Net trained with five class criteria—derived from previous classification criteria—yielded valid results (overall accuracy:65.6 %). The most accurate results were acquired from trained U-Net with all input datasets; in particular, the tidal channel density caused a significant increase in accuracy. The accuracy of the U-Net was approximately 20 % higher than that of other classification models. These results demonstrate that surface sediment classification using UAV images and the U-Net model is effective.

Improving field-scale crop actual evapotranspiration monitoring with Sentinel-3, Sentinel-2, and Landsat data fusion

One of the primary applications of satellite Land Surface Temperature (LST) observations lies in their utilization for modeling of actual evapotranspiration (ET) in agricultural crops, with the primary goals of monitoring and enhancing irrigation practices and improving crop water use productivity, as stipulated by Sustainable Development Goal (SDG) indicator 6.4.1. Evapotranspiration is a complex and dynamic process, both temporally and spatially, necessitating LST observations with high spatio-temporal resolution. Presently, none of the existing spaceborne thermal sensors can provide quasi-daily field-scale LST observations, prompting the development of methods for data fusion (thermal sharpening) of observations from various shortwave and thermal sensors to meet this spatio-temporal requirement. Previous research has demonstrated the effectiveness of combining shortwave-multispectral Sentinel-2 observations with thermal-infrared Sentinel-3 observations to derive daily, field-scale LST and ET estimates. However, these studies also highlighted limitations in capturing the distinct thermal contrast between cooler LST in irrigated agricultural areas and the hotter, adjacent dry regions. In this study, we aim to address this limitation by incorporating information on thermal spatial variability observed by Landsat satellites into the data fusion process, without being constrained by infrequent or cloudy Landsat thermal observations and while retaining the longwave radiance emission captured by the Sentinel-3 thermal sensor at its native resolution. Two approaches are evaluated, both individually and as a complementary combination, and validated against in situ LST measurements. The best performing approach, which leads to reduction in root mean square error of up to 1.5 K when compared to previous research, is subsequently used to estimate parcel-level actual evapotranspiration. The ET modeling process has also undergone various improvements regarding the gap-filling of input and output data, input datasets and code implementation. The resulting ET is validated using lysimeters and eddy covariance towers in Spain, Lebanon, Tunisia, and Senegal resulting in minimal overall bias (systematic underestimation of less than 0.07 mm/day) and a low root mean square error (down to 0.84 mm/day) when using fully global input datasets. The enhanced LST sharpening methodology is sensor agnostic and should remain relevant for the upcoming thermal missions while the accuracy of the modeled ET fluxes is encouraging for further utilization of observations from Sentinel satellites, and other Copernicus data, for monitoring SDG indicator 6.4.1.