Input Dataset Research Articles

COWS for High Performance: Cost Aware Work Stealing for Irregular Parallel Loop

Parallel libraries such as OpenMP distribute the iterations of parallel-for-loops among the threads, using a programmer-specified scheduling policy. While the existing scheduling policies perform reasonably well in the context of balanced workloads, in computations that involve highly imbalanced workloads it is extremely non-trivial to obtain an efficient distribution of work (even using non-static scheduling methods like dynamic and guided). In this paper, we present a scheme called COst aware Work Stealing (COWS) to efficiently extend the idea of work-stealing to OpenMP. In contrast to the traditional work-stealing schedulers, COWS takes into consideration that (i) not all iterations of a parallel-for-loops may take the same amount of time. (ii) identifying a suitable victim for stealing is important for load-balancing, and (iii) queues lead to significant overheads in traditional work-stealing and should be avoided. We present two variations of COWS: WSRI (a naive work-stealing scheme based on the number of remaining iterations) and WSRW (work-stealing scheme based on the amount of remaining workload). Since in irregular loops like those found in graph analytics it is not possible to statically compute the cost of the iterations of the parallel-for-loops, we use a combined compile-time + runtime approach, where the remaining workload of a loop is computed efficiently at runtime by utilizing the code generated by our compile-time component. We have performed an evaluation over seven different benchmark programs, using five different input datasets, on two different hardware across a varying number of threads; leading to a total number of 275 configurations. We show that in 225 out of 275 configurations, compared to the best OpenMP scheduling scheme for that configuration, our approach achieves clear performance gains.

Gynaecological Artificial Intelligence Diagnostics (GAID) GAID and Its Performance as a Tool for the Specialist Doctor.

Human-centric artificial intelligence (HCAI) aims to provide support systems that can act as peer companions to an expert in a specific domain, by simulating their way of thinking and decision-making in solving real-life problems. The gynaecological artificial intelligence diagnostics (GAID) assistant is such a system. Based on artificial intelligence (AI) argumentation technology, it was developed to incorporate, as much as possible, a complete representation of the medical knowledge in gynaecology and to become a real-life tool that will practically enhance the quality of healthcare services and reduce stress for the clinician. Our study aimed to evaluate GAIDS' efficacy and accuracy in assisting the working expert gynaecologist during day-to-day clinical practice. Knowledge-based systems utilize a knowledge base (theory) which holds evidence-based rules ("IF-THEN" statements) that are used to prove whether a conclusion (such as a disease, medication or treatment) is possible or not, given a set of input data. This approach uses argumentation frameworks, where rules act as claims that support a specific decision (arguments) and argue for its dominance over others. The result is a set of admissible arguments which support the final decision and explain its cause. Based on seven different subcategories of gynaecological presentations-bleeding, endocrinology, cancer, pelvic pain, urogynaecology, sexually transmitted infections and vulva pathology in fifty patients-GAID demonstrates an average overall closeness accuracy of zero point eighty-seven. Since the system provides explanations for supporting a diagnosis against other possible diseases, this evaluation process further allowed for a learning process of modular improvement in the system of the diagnostic discrepancies between the system and the specialist. GAID successfully demonstrates an average accuracy of zero point eighty-seven when measuring the closeness of the system's diagnosis to that of the senior consultant. The system further provides meaningful and helpful explanations for its diagnoses that can help clinicians to develop an increasing level of trust towards the system. It also provides a practical database, which can be used as a structured history-taking assistant and a friendly, patient record-keeper, while improving precision by providing a full list of differential diagnoses. Importantly, the design and implementation of the system facilitates its continuous development with a set methodology that allows minimal revision of the system in the face of new information. Further large-scale studies are required to evaluate GAID more thoroughly and to identify its limiting boundaries.

Predicting Day-Ahead Electricity Market Prices through the Integration of Macroeconomic Factors and Machine Learning Techniques

Several events in the last years changed to some extent the common understanding of the electricity day-ahead market (DAM). The shape of the electricity price curve has been altered as some factors that underpinned the electricity price forecast (EPF) lost their importance and new influential factors emerged. In this paper, we aim to showcase the changes in EPF, understand the effects of uncertainties and propose a forecasting method using machine learning (ML) algorithms to cope with random events such as COVID-19 pandemic and the conflict in Black Sea region. By adjusting the training period according to the standard deviation that reflects the price volatility, feature engineering and by using two regressors for weighing the results, significant improvements in the performance of the EPF are achieved. One of the contributions of the proposed method consists in adjusting the training period considering the price variation. Thus, we introduce a rule-based approach given an empirical observation that for days with a higher growth in prices the training interval should be shortened, capturing the sharp variations of prices. The results of several cutting-edge ML algorithms represent the input for a predictive meta-model to obtain the best forecasting solution. The input dataset spans from Jan. 2019 to Aug. 2022, testing the proposed EPF method for both stable and more tumultuous intervals and proving its robustness. This analysis provides decision makers with an understanding of the price trends and suggests measures to combat spikes. Numerical findings indicate that on average mean absolute error (MAE) improved by 48% and root mean squared error (RMSE) improved by 44% compared to the baseline model (without feature engineering/adjusting training). When the output of the ML algorithms is weighted using the proposed meta-model, MAE further improved by 2.3% in 2020 and 5.14% in 2022. Less errors are recorded in stable years like 2019 and 2020 (MAE = 6.71, RMSE = 14.67) compared to 2021 and 2022 (MAE = 9.45, RMSE = 20.64).

Geomorphometric maps of Australia’s Marine Park estate and their role in improving the integrated monitoring and management of marine ecosystems

The loss of marine biodiversity is a major global issue that needs to be prioritised. In Australia, a considerable proportion (48%) of its Exclusive Economic Zone is dedicated to marine protected areas. To effectively manage this network of marine protected areas Australia has recently introduced a Management Effectiveness system. This system is designed to identify, monitor, and manage natural values and the various activities and pressures affecting the Australian Marine Parks (AMPs). Key to this approach is the identification and accurate mapping of the location of these values and pressures acting on the seabed. The AusSeabed program is a national initiative in Australia aimed at improving access to bathymetric data and coordinating efforts to collect such data in Australian waters. This manuscript proposes a novel systematic processing method to create detailed and scalable geomorphometric maps from AusSeabed’s bathymetric data holdings, intended as a standard operating procedure for initial bathymetric data interpretation in the AMPs. Utilising this workflow, we produce seafloor geomorphometry maps across 37 AMPs within which sufficient bathymetric data has been collected. These maps can be used 1) for predictive mapping of biological assemblages; 2) in field sampling design for the collection of ‘ground truthing’ data (e.g. underwater imagery and sediment samples) to validate habitat maps from bathymetric datasets; and 3) as input datasets for subsequent geomorphological mapping with a deeper understanding of seafloor processes. This research highlights the importance of robust geomorphometry classification standards to ensure consistency in mapping Australia’s marine estate in preparation for the Decade of Oceans plans. The Seamap Australia program provides a stepwise approach to advancing Australia’s national collection of bathymetric data into derived products that can enable habitat mapping of Australian waters, providing a foundational tool for the adaptive management of AMPs.

Simulation-based optimization of workforce configuration for multi-division maintenance departments

The configuration of human resources in a production company that determines the number of employees at different technical or nontechnical competencies engaged in various operational tasks should be constructed to meet the task requirements above pre-specified expectation levels. Human resource requirements show variation depending on the divisional structures within an organization and their operational dependencies and priorities. A maintenance facility is generally recognized as one of the most labor-intensive divisions in any production company. It employs several workers with diverse technical skills to ensure machinery reliability and availability targets in line with the production plans. The current research study aims to develop a continuous-event simulation algorithm to optimize the maintenance crew configuration and allocation decisions for production companies where different clusters of failure modes are available for multi-equipment operations. This algorithm also simultaneously evaluates the reallocation of the occupied crew for other overlapping maintenance tasks that require similar technical competency. Accordingly, each maintenance division with varying personnel numbers is authorized to maintain particular failure modes. The developed model was implemented for the maintenance department of a surface coal mine where five excavators exposed to diversified random failures are employed. The model input dataset covers three years of maintenance records classified into mechanical and electrical failure modes. The simulation outcomes reveal that the cumulative indirect and direct costs can be minimized for a crew configuration including six electricians and four mechanics.

Time analysis of online consumer behavior by decision trees, GUHA association rules, and formal concept analysis

AbstractData analytics plays a significant role within the context of the digital business landscape, particularly concerning online sales, aiming to enhance understanding of customer behaviors in the online realm. We review the recent perspectives and empirical findings from several years of scholarly investigation. Furthermore, we propose combining computational methods to scrutinize online customer behavior. We apply the decision tree construction, GUHA (General Unary Hypotheses Automaton) association rules, and Formal concept analysis for the input dataset of 9123 orders (transactions) of sports nutrition, healthy foods, fitness clothing, and accessories. Data from 2014 to 2021, covering eight years, are employed. We present the empirical discoveries, engage in a critical discourse concerning these findings, and delineate the constraints inherent in the research process. The decision tree for classification of the year’s fourth quarter implies that the most important attributes are country, gross profit category, and delivery. The classification of the morning time implies that the most important attributes are gender and country. Thus, the potential marketing strategies can include heterogeneous conditions for men and women based on these findings. Analyzing the identified groups of customers by concept lattices and GUHA association rules can be valuable for targeted marketing, personalized recommendations, or understanding customer preferences.

Modeling population distribution: A visual and quantitative analysis of gradient boosting and deep learning models for multi‐output spatial disaggregation

AbstractSpatially aggregated data on socio‐demographic groups often fail to capture the population's spatial heterogeneity in cities. This poses challenges for urban planning, particularly when addressing the needs of groups such as migrants or families with children. Moreover, the commonly provided aggregated units, such as census tracts, vary in size and across data sources. Existing literature on disaggregation typically handles individual subgroups separately, ignoring their interrelations in the downscaling process. This article explores the potentials of multi‐output regression models for simultaneous spatial downscaling of multiple groups and conducts a detailed spatial error analysis using individualized neighborhoods. We experiment with self‐training gradient‐boosting trees and fully convolutional neural networks, assessing the quality of results against ground truth data at the target resolution. We show that the evaluation of the disaggregated results at this detailed resolution requires unconventional methods. The methodology proves convenient and achieves high‐accuracy results using input datasets of building features.

SHIP: a computational framework for simulating and validating novel technologies in hardware spiking neural networks.

Investigations in the field of spiking neural networks (SNNs) encompass diverse, yet overlapping, scientific disciplines. Examples range from purely neuroscientific investigations, researches on computational aspects of neuroscience, or applicative-oriented studies aiming to improve SNNs performance or to develop artificial hardware counterparts. However, the simulation of SNNs is a complex task that can not be adequately addressed with a single platform applicable to all scenarios. The optimization of a simulation environment to meet specific metrics often entails compromises in other aspects. This computational challenge has led to an apparent dichotomy of approaches, with model-driven algorithms dedicated to the detailed simulation of biological networks, and data-driven algorithms designed for efficient processing of large input datasets. Nevertheless, material scientists, device physicists, and neuromorphic engineers who develop new technologies for spiking neuromorphic hardware solutions would find benefit in a simulation environment that borrows aspects from both approaches, thus facilitating modeling, analysis, and training of prospective SNN systems. This manuscript explores the numerical challenges deriving from the simulation of spiking neural networks, and introduces SHIP, Spiking (neural network) Hardware In PyTorch, a numerical tool that supports the investigation and/or validation of materials, devices, small circuit blocks within SNN architectures. SHIP facilitates the algorithmic definition of the models for the components of a network, the monitoring of states and output of the modeled systems, and the training of the synaptic weights of the network, by way of user-defined unsupervised learning rules or supervised training techniques derived from conventional machine learning. SHIP offers a valuable tool for researchers and developers in the field of hardware-based spiking neural networks, enabling efficient simulation and validation of novel technologies.

Isotope ratios of total C, N, and S in particulate matter simultaneously calibrated by mixed USGS and IAEA reference materials

This study presents a method for calibrating the isotope ratios of the total carbon, nitrogen, and sulfur in particulate matter (PM) collected from the Seoul metro using an elemental analyzer-isotope ratio mass spectrometer (EA-IRMS). Mixtures of isotope reference materials (MRMs) from the U.S. Geological Survey (USGS) and International Atomic Energy Agency (IAEA) reference materials formed an input dataset for generalized least squares (GLS) regression to yield calibration lines. The analytical method proposed in this study enabled the measurement of stable isotope ratios of total carbon, nitrogen, and sulfur simultaneously. Results showed good linearity and repeatability for carbon and nitrogen isotopes, but poor results for sulfur isotopes due to peak broadening. Reference values with uncertainties for the isotope ratios of total carbon, nitrogen, and sulfur were determined for the collected PM, demonstrating twice as much uncertainty as that of the USGS and IAEA reference materials. Homogeneity was the biggest uncertainty source for the calibrated values.

MK-BMC: a Multi-Kernel framework with Boosted distance metrics for Microbiome data for Classification.

Research on human microbiome has suggested associations with human health, opening opportunities to predict health outcomes using microbiome. Studies have also suggested that diverse forms of taxa such as rare taxa that are evolutionally related and abundant taxa that are evolutionally unrelated could be associated with or predictive of a health outcome. Although prediction models were developed for microbiome data, no prediction models currently exist that use multiple forms of microbiome-outcome associations. We developed MK-BMC, a Multi-Kernel framework with Boosted distance Metrics for Classification using microbiome data. We propose to first boost widely used distance metrics for microbiome data using taxon-level association signal strengths to up-weight taxa that are potentially associated with an outcome of interest. We then propose a multi-kernel prediction model with one kernel capturing one form of association between taxa and the outcome, where a kernel measures similarities of microbiome compositions between pairs of samples being transformed from a proposed boosted distance metric. We demonstrated superior prediction performance of (i) boosted distance metrics for microbiome data over original ones and (ii) MK-BMC over competing methods through extensive simulations. We applied MK-BMC to predict thyroid, obesity, and inflammatory bowel disease status using gut microbiome data from the American Gut Project and observed much-improved prediction performance over that of competing methods. The learned kernel weights help us understand contributions of individual microbiome signal forms nicely. Source code together with a sample input dataset is available at https://github.com/HXu06/MK-BMC.

Critical analysis of hybrid learning models to detect morphed images

Machine learning (ML) algorithms produce different results with the kind of input data set or parameters passed to the algorithm. As per input data set, the analysis generally depends on the data type and amount of data set being used to train and test. Other parameters which affect the machine learning analysis are the CNN parameters or the hybrid approach being used. As there is no similar result for ML analysis on various data set, it is hard to predict which model will work most efficiently on the given data set. In this paper an analytical analysis of machine learning algorithms has been done to examine the working of various algorithms individually or in Hybrid mode. This paper studies CNN, CNN + RF, CNN +SVM approach and the theoretical parameters affecting them which helps in deciding the best suited algorithm with the given data set.

Physical consistency and invariance in machine learning of turbulent signals

This paper concerns an investigation of the invariance and consistency of deep learning of turbulent pressure fluctuations. The long-short-memory model is employed to predict wall pressure fluctuations across physical regimes featuring turbulence, shock–boundary layer interaction, and separation. The model's sensitivity to the data inputs is examined using different input data sets. Training the deep learning model based on the raw signals from different flow regions leads to large inaccuracies. It is shown that the data must be appropriately pre-processed before training for the deep learning model predictions to become consistent. Removing the mean and using the normalized fluctuating component of the signal, the deep learning predictions not only greatly improved in accuracy but, most importantly, converged and became consistent, provided that the signal sparsity remains within the inertial sub-range of the turbulence energy spectrum cascade. The power spectra of the surface pressure fluctuations reveal that the model provides high accuracy up to a certain frequency for the fully turbulent flow. The deep learning model's consistency is evidenced by being transferable across the various probe positions on the wall despite the significant differences in the turbulent flow properties in the training data set, i.e., signals obtained before, after, and inside the shock–boundary layer interaction regions. The model's prediction consistency and invariance to the turbulent signal training location(s) are promising for applying deep learning models to various turbulent flows.

Scalable Hypergraph Visualization.

Hypergraph visualization has many applications in network data analysis. Recently, a polygon-based representation for hypergraphs has been proposed with demonstrated benefits. However, the polygon-based layout often suffers from excessive self-intersections when the input dataset is relatively large. In this paper, we propose a framework in which the hypergraph is iteratively simplified through a set of atomic operations. Then, the layout of the simplest hypergraph is optimized and used as the foundation for a reverse process that brings the simplest hypergraph back to the original one, but with an improved layout. At the core of our approach is the set of atomic simplification operations and an operation priority measure to guide the simplification process. In addition, we introduce necessary definitions and conditions for hypergraph planarity within the polygon representation. We extend our approach to handle simultaneous simplification and layout optimization for both the hypergraph and its dual. We demonstrate the utility of our approach with datasets from a number of real-world applications.

Prediction of the Dynamic Properties of Concrete Using Artificial Neural Networks

This study explores how dynamic characteristics of concrete, such as dynamic shear modulus, dynamic modulus of elasticity, and dynamic Poisson's ratio, affect stability and performance in civil engineering applications. Traditional testing procedures, which include the time-consuming and costly process of mixing and casting specimens, are both time-consuming and costly. The primary objective of this research is to improve efficiency by using Artificial Neural Networks (ANNs) and regression analysis to predict the dynamic properties of concrete, providing a machine-learning-based alternative to traditional experimental methodologies. A set of 72 concrete specimens was methodically built and evaluated, with compressive strengths of 50 MPa, aspect ratios ranging from 1 to 2.5, and an average density of 2400 kg/m3. An input dataset and ANN targets were built using these samples. The ANN model, which used cutting-edge deep learning techniques, went through extensive training, validation, and testing, as well as statistical regression analysis. A comparison shows that the predicted dynamic modulus of elasticity and shear modulus using both ANN and regression approaches nearly match the experimental values, with a maximum error of 5%. Despite good forecasts for the dynamic Poisson's ratio, errors of up to 20% were detected on occasion, which were attributed to sample shape variations. Doi: 10.28991/CEJ-2024-010-01-016 Full Text: PDF

Comparison of soft-computing techniques: Data-driven models for flood forecasting

<p>Accurate flood forecasting is a crucial process for predicting the timing, occurrence, duration, and magnitude of floods in specific zones. This prediction often involves analyzing various hydrological, meteorological, and environmental parameters. In recent years, several soft computing techniques have been widely used for flood forecasting. In this study, flood forecasting for the Narmada River at the Hoshangabad gauging site in Madhya Pradesh, India, was conducted using an Artificial Neural Network (ANN) model, a Fuzzy Logic (FL) model, and an Adaptive Neuro-Fuzzy Inference System (ANFIS) model. To assess their capacity to handle different levels of information, three separate input data sets were used. Our objective was to compare the performance and evaluate the suitability of soft computing data-driven models for flood forecasting. For the development of these models, monthly discharge data spanning 33 years from six gauging sites were selected. Various performance measures, such as regression, root mean square error (RMSE), and percentage deviation, were used to compare and evaluate the performances of the different models. The results indicated that the ANN and ANFIS models performed similarly in some cases. However, the ANFIS model generally predicted much better than the ANN model in most cases. The ANFIS model, developed using the hybrid method, delivered the best performance with an RMSE of 211.97 and a coefficient of regression of 0.96, demonstrating the potential of using these models for flood forecasting. This research highlighted the effectiveness of soft computing techniques in flood forecasting and established useful suitability criteria that can be employed by flood control departments in various countries, regions, and states for accurate flood prognosis.</p>

PUTS: Privacy-Preserving and Utility-Enhancing Framework for Trajectory Synthesization

Vehicle trajectory data is essential for traffic management and location-based services. However, publishing real-life trajectory data has been challenging because vehicle trajectories contain users' sensitive information. Differential privacy addresses such problems by publishing a synthetic version of the input dataset, but existing works always assume the real-world data is absolutely accurate. This assumption no longer holds in trajectory data because it typically contains errors due to inaccurate positioning services, which leads to poor performance of data synthesized by such trajectories. Even worse, existing works may generate unrealistic trajectories due to their coarse data synthesis methods, resulting in low practical utility or even inability to handle complex tasks. In this paper, we propose a <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">P</u> rivacy-preserving and <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">U</u> tility-enhancing framework for <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">T</u> rajectory <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S</u> ynthesization ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PUTS</i> ). Our framework mitigates the impact of data errors in trajectories on differential privacy mechanisms, by exploiting map-matching techniques and real-world road network structure. In <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PUTS</i> , a two-layer approach from path to trajectory synthesis is proposed to not only guarantee the reality of synthetic trajectories, but also scale up <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PUTS</i> in real-world applications. Extensive experiments on real-world datasets show that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PUTS</i> significantly outperforms existing methods in terms of utility in a range of real-world applications.

Hybrid arithmetic optimization algorithm with deep learning model for secure Unmanned Aerial Vehicle networks

&lt;abstract&gt; &lt;p&gt;Securing Unmanned Aerial Vehicle (UAV) systems is vital to safeguard the processes involved in operating the drones. This involves the execution of robust communication encryption processes to defend the data exchanged between the UAVs and ground control stations. Intrusion detection, powered by Deep Learning (DL) techniques such as Convolutional Neural Networks (CNN), allows the classification and identification of potential attacks or illegal objects in the operational region of the drone, thus distinguishing them from the routine basics. The current research work offers a new Hybrid Arithmetic Optimizer Algorithm with DL method for Secure Unmanned Aerial Vehicle Network (HAOADL-UAVN) model. The purpose of the proposed HAOADL-UAVN technique is to secure the communication that occurs in UAV networks via threat detection. At the primary level, the network data is normalized through min-max normalization approach in order to scale the input dataset into a useful format. The HAOA is used to select a set of optimal features. Next, the security is attained via Deep Belief Network Autoencoder (DBN-AE)-based threat detection. At last, the hyperparameter choice of the DBN-AE method is implemented using the Seagull Optimization Algorithm (SOA). A huge array of simulations was conducted using the benchmark datasets to demonstrate the improved performance of the proposed HAOADL-UAVN algorithm. The comprehensive results underline the supremacy of the HAOADL-UAVN methodology under distinct evaluation metrics.&lt;/p&gt; &lt;/abstract&gt;

Aerosol source apportionment uncertainty linked to the choice of input chemical components

For a Positive Matrix Factorization (PMF) aerosol source apportionment (SA) studies there is no standard procedure to select the most appropriate chemical components to be included in the input dataset for a given site typology, nor specific recommendations in this direction. However, these choices are crucial for the final SA outputs not only in terms of number of sources identified but also, and consequently, in the source contributions estimates. In fact, PMF tends to reproduce most of PM mass measured independently and introduced as a total variable in the input data, regardless of the percentage of PM mass which has been chemically characterized, so that the lack of some specific source tracers (e.g. levoglucosan) can potentially affect the results of the whole source apportionment study. The present study elaborates further on the same concept, evaluating quantitatively the impact of lacking specific sources’ tracers on the whole source apportionment, both in terms of identified sources and source contributions. This work aims to provide first recommendations on the most suitable and critical components to be included in PMF analyses in order to reduce PMF output uncertainty as much as possible, and better represent the most commons PM sources observed in many sites in Western countries. To this aim, we performed three sensitivity analyses on three different datasets across EU, including extended sets of organic tracers, in order to cover different types of urban conditions (Mediterranean, Continental, and Alpine), source types, and PM fractions. Our findings reveal that the vehicle exhaust source resulted to be less sensitive to the choice of analytes, although source contributions estimates can deviate significantly up to 44 %. On the other hand, for the detection of the non-exhaust one is clearly necessary to analyze specific inorganic elements. The choice of not analysing non-polar organics likely causes the loss of separation of exhaust and non-exhaust factors, thus obtaining a unique road traffic source, which provokes a significant bias of total contribution. Levoglucosan was, in most cases, crucial to identify biomass burning contributions in Milan and in Barcelona, in spite of the presence of PAHs in Barcelona, while for the case of Grenoble, even discarding levoglucosan, the presence of PAHs allowed identifying the BB factor. Modifying the rest of analytes provoke a systematic underestimation of biomass burning source contributions. SIA factors resulted to be generally overestimated with respect to the base case analysis, also in the case that ions were not included in the PMF analysis. Trace elements were crucial to identify shipping emissions (V and Ni) and industrial sources (Pb, Ni, Br, Zn, Mn, Cd and As). When changing the rest of input variables, the uncertainty was narrow for shipping but large for industrial processes. Major and trace elements were also crucial to identify the mineral/soil factor at all cities. Biogenic SOA and Anthropogenic SOA factors were sensitive to the presence of their molecular tracers, since the availability of OC alone is unable to separate a SOA factor. Arabitol and sorbitol were crucial to detecting fungal spores while odd number of higher alkanes (C27 to C31) for plant debris.

Extracting Consumers’ Perceptions for Indonesian Spice Drinks Using Social Media Data Mining and Kansei Engineering

Local factors and global influences shape consumers’ perceptions through social media. In this regard, spice drinks as an agribusiness product have received increasing attention due to the Covid-19 pandemic. Therefore, understanding consumers’ perceptions is crucial for promoting the development of spice drinks. This study aims to (1) extract consumers’ perceptions of spice drinks based on discussions on social media using sentiment analysis and (2) classify the factors influencing their perceptions using factor analysis. The input dataset was obtained through Twitter data to acquire Kansei words. The results disclosed that Twitter could extract Kansei words and validate consumers’ perceptions of spice drinks as an agribusiness product. The sentiment analysis revealed 78% positive and 13% neutral tweets. Subsequently, an online survey was conducted among 495 respondents aged 18 to 41, distributed through various social media platforms from June to August 2022. The respondents were Generation Z and Millennials, with Generation Z referring to individuals born between 1997 and 2012 and Millennials born between 1981 and 1996. Factor analysis identified four principal components influencing consumers’ perceptions toward spice drinks: positive attitudes were associated with the quick, milky, mixed, healthy, quality, energy, fresh, warm, and safe; benefits were affiliated with the words enjoy, rest, life, smile, and story; quality concerned easy, flavour, and spicy; and sensory dealt with sweet, aroma, and bitter.

PREDICTION SYSTEM FOR THE INFLUENCE OF PHOSPHORUS ON WHEAT YIELD: OPTIMAL HYPERPARAMETER SELECTION

Modern studies in the application of machine learning in agrotechnology have showcased significant advancements in accurate crop yield forecasting and agricultural production optimization. This study aimed to select optimal hyperparameters for a neural network designed to forecast the impact of phosphorus on the yield of spring wheat. Tasks to achieve this goal included the selection of optimal neural network hyperparameters and deployment of the prediction model into forecasting system, ensuring accessibility and user convenience. Input dataset consisted of year, region, climatic indicators (temperature of the soil surface, precipitation, humidity), applied phosphorus, and the target variable is the yield of spring wheat. To select the optimal hyperparameters, GridSearchCV was utilized and integrated into the experiment, enabling the achievement of more precise and impactful forecasting outcomes. Experiments aimed at hyperparameter tuning identified an optimal network configuration (3 layers, 32 neurons, 300 epochs, 32 batches). Results from training showcased that this network configuration achieved the best performance, demonstrating minimal mean squared error (MSE). These results introduced a novel scientific approach in training data using this neural network model for forecasting the yield of cereal crops. Developing the prediction system's interface relied on the Streamlit, allowing the creation of an intuitive and user-friendly interface. The visualization derived from this model empowers users to evaluate crop yield and receive recommendations for optimal phosphorus application to enhance spring wheat yield. These results highlighted practical significance of this prediction system, based on neural network utilizing optimal hyperparameters that are ready for practical implementation by system developers.