Machine Learning Capabilities Research Articles

APPLYING MACHINE LEARNING IN THE U.S. POLITICAL LANDSCAPE: FORECASTING, POLLING, AND THE DOMESTIC SUPPLY CHAIN

There are areas of the current U.S. political system that are fit for the implementation of machine learning capabilities. Some of these areas include election forecasting, polling, and vote-by-mail services. The introduction of machine learning tools to the components of the U.S. political system can result in increases in efficiency and accessibility throughout the sector. By doing so, there would be an assumed decrease in necessary labor. This research analyzed the specified areas of U.S. politics by establishing a four-step framework based on a previous exploratory case study into the application of machine learning to operations management and adjusted said framework to fit the needs of the nature of U.S. politics. This framework was utilized to determine the objectives to which machine learning would be applied within this sector. As well, it serves to guide the technology or strategies necessary to achieve those objectives, illustrate the effect on performance and on stakeholders. This analysis introduces valuable insights for any decision makers as well as providing an accessible and informative review of some of the potential applications of machine learning in the existing U.S. political system.

Assessment of the Accuracy of Various Machine Learning Algorithms for Classifying Urban Areas through Google Earth Engine: A Case Study of Kabul City, Afghanistan

Accurate identification of urban land use and land cover (LULC) is important for successful urban planning and management. Although previous studies have explored the capabilities of machine learning (ML) algorithms for mapping urban LULC, identifying the best algorithm for extracting specific LULC classes in different time periods and locations remains a challenge. In this research, three machine learning algorithms were employed on a cloud-based system to categorize urban land use of Kabul city through satellite images from Landsat-8 and Sentinel-2 taken in 2023. The most advanced method of generating accurate and informative LULC maps from various satellite data and presenting accurate outcomes is the machine learning algorithm in Google Earth Engine (GEE). The objective of the research was to assess the precision and efficiency of various machine learning techniques, such as random forest (RF), support vector machine (SVM), and classification and regression tree (CART), in producing dependable LULC maps for urban regions by analyzing optical satellite images of sentinel and Landsat taken in 2023. The urban area was divided into five classes: built-up area, vegetation, bare-land, soil, and water bodies. The accuracy and validation of all three algorithms were evaluated. The RF classifier showed the highest overall accuracy of 93.99% and 94.42% for Landsat-8 and Sentinel-2, respectively, while SVM and CART had lower overall accuracies of 87.02%, 81.12%, and 91.52%, 87.77%, with Landsat-8 and Sentinel-2, respectively. The results of the present study revealed that in this classification and comparison, RF performed better than SVM and CART for classifying urban territory for Landsat-8 and Sentinel-2 using GEE. Furthermore, the study highlights the importance of comparing the performance of different algorithms before selecting one and suggests that using multiple methods simultaneously can lead to the most precise map.

Interpretable machine learning on large samples for supporting runoff estimation in ungauged basins

The distribution of flowmeter data and basin characteristic information exhibits substantial disparities, with most flow observations being recorded at a limited number of well-monitored locations. The perennial challenge of achieving reliable and robust hydrological modeling in ungauged catchments through regionalization has persisted. The increasing availability of large-scale hydrological datasets, coupled with recent advancements in machine learning techniques, offers new opportunities to explore patterns of association between basin attributes and hydrological parameters to enhance streamflow predictions. A novel parameter cross-regional transfer approach based on interpretable machine learning (XGBoost) is proposed to accurately predict runoff processes in ungauged regions by leveraging well-trained models across numerous basins within climate zones. We validate the effectiveness of this framework across 5,764 basins in a large sample dataset (Caravan), employing Nash-Sutcliffe Efficiency (NSE), RMSE and bias to assess performance. And a comparison is made with deep transfer learning based on LSTM and Transformer. Results indicate that the proposed method achieves NSE values exceeding 0.2 for 75 % of the ungauged basins, demonstrating superior performance and more stable accuracy compared to pure deep learning models, owing to its incorporation of physical constraints. Furthermore, the response of parameters to basin attributes within different climatic zones in the large-sample context is elucidated through SHAP values, enriching the understanding of hydrological features through data-driven inverse inference. These findings underscore the capability of interpretable machine learning to leverage hydro-physical regularities extracted from abundant basin features, thereby enhancing the accuracy of runoff predictions in ungauged regions.

ЦИФРОВІ ТРАНСФОРМАЦІЇ ПРОГРАМ ЛОЯЛЬНОСТІ ПІДПРИЄМСТВ ТОРГІВЛІ

The development of retail, due to the emergence of new modern digital tools, is changing approaches to ensuring and managing consumer loyalty within the framework of the implementation of the concept of relationship marketing. Loyalty programs implemented by retail enterprises need to be transformed to ensure interaction with consumers in all available channels using the capabilities of artificial intelligence, machine learning and Big Data. The purpose of the article is to determine the prerequisites and directions of transformation of loyalty programs in accordance with the possibilities of using digital tools relevant for retail to ensure interaction with consumers within the omnichannel approach. The object of the research is loyalty programs of retail trade enterprises as a set of tools aimed at strengthening the positive experience of interaction with consumers and increasing their level of satisfaction with the offered services. The methods used in the research. Here are used methods of categorical analysis and synthesis to study theoretical approaches to determining the essence and features of omnichannel loyalty programs. In addition, the description of the transformation of loyalty programs required the use of the observation method to display market trends and vectors of their influence on the specified process, as well as specification - to highlight and characterize relevant digital tools. Research hypothesis. The development of technologies and the emergence of new digital tools in retail, along with changes in purchasing consumers’ behavior, necessitates the transformation of loyalty programs. The expediency of interacting with consumers at all possible touch points determines the relevance of an omnichannel loyalty program. Presenting main material. In order to form consumer loyalty, trade enterprises must ensure a high level of their satisfaction with the service process, which is associated with the use of influence opportunities with the help of special programs. Factors in the marketing environment change consumer requirements for loyalty programs, the use of new digital tools in retail allows for effective interaction within the scope of the omnichannel approach. The originality and practical significance of the research lies in proving the relevance of the transformation of loyalty programs using digital tools. Conclusions and perspectives of further research. The omnichannel loyalty program allows to increase the efficiency of retail trade enterprises. Digital tools implemented in omnichannel loyalty programs and the specifics of their application, in our opinion, are the subject of separate researches.

Machine Learning and Possibility of its Application for Forensic Researches on Firearms

Role of machine learning in modern society cannot be over-estimated, especially given its steadily increasing importance in recent years. Various machine learning models are applied not only in everyday life but also in solving practical, engineering, and scientific challenges. However, the methods and capabilities of machine learning remain relatively obscure to most professionals across various fields of knowledge, including specialists in forensic firearm examination. The article discusses basic tasks that machine learning methods can facilitate and provides examples of tasks faced by forensic experts during firearm examinations that can be solved utilizing these methods. The article purpose is to familiarize professionals in the field of forensic science with be-sic capabilities of machine learning methods. The research paper demonstrates which specific types of tasks in forensic firearms examination it is advisable to apply these methods to, and what is required for this purpose. In order to achieve this goal, general scientific methods (dialectical, comparative, analysis, synthesis, induction, deduction) and special scientific methods (formal-logical, systemic-structural) have been employed. Specific expertise and skills in forensic identification methods (including ballistics, microscopic comparative analysis of traces on bullets and casings, detection and examination of gunshot residue to estimate shooting distance, etc.), principles of machine learning, and mathe-matical algorithms used to construct various models have been applied.

Employing machine learning techniques for estimating the differential code biases of GPS satellites

ABSTRACT In this study, the capabilities of Machine Learning (ML) are exploited to predict the Differential Code Biases (DCBs) of Global Positioning System (GPS) satellites from the broadcast Total Group Delays (TGDs), satellite numbers, satellite block types, and Sunspot numbers. Firstly, a detailed analysis of the DCB and TGD values over five years is provided. Then, different ML models are trained and tested. The results showed that the bagged trees, the rational quadratic Gaussian Process Regression, and the Feed-Forward Neural Networks (2 hidden layers) models can be used efficiently to predict the DCB values of GPS satellites.

Artificial Intelligence Could be the Personalized Treatment Strategy for Cancer

Cancer is one of the world’s most serious medical challenges. Because of its great heterogeneity, persons with similar tumors could react in a different way to the same drugs or surgical methods, prompting the development of more accurate tumor treatment approaches as well as patient-specific tailored treatments. To establish targeted therapy choices for patients, it is critical to have a full acceptance of the changes that tumors endure, counting modifications in their genetic factors, proteins, and cancer cell behaviors. Tumor treatment requires precise targeting. Big data-driven artificial intelligence (AI) may reveal patterns, insights, and related information hidden inside huge volumes of data. To identify exact data from transcriptomics, radiomics, genomes, proteomics, digital pathological images, etc., subsets of AI’s machine learning capability may be explored. This may help clinicians get a better and more comprehensive knowledge of malignancies. In addition, to provides the optimal therapy for each patient and improves clinical outcomes.

Prediction and reliability analysis of shear strength of RC deep beams

This study explores machine learning (ML) capabilities for predicting the shear strength of reinforced concrete deep beams (RCDBs). For this purpose, eight typical machine-learning models, i.e., symbolic regression (SR), XGBoost (XGB), CatBoost (CATB), random forest (RF), LightGBM, support vector regression (SVR), artificial neural networks (ANN), and Gaussian process regression (GPR) models, are selected and compared based on a database of 840 samples with 14 input features. The hyperparameter tuning of the introduced ML models is performed using the Bayesian optimization (BO) technique. The comparison results show that the CatBoost model is the most reliable and accurate ML model (R2 = 0.997 and 0.947 in the training and testing sets, respectively). In addition, simple and practical design expressions for RCDBs have been proposed based on the SR model with a physical meaning and acceptable accuracy (an average prediction-to-test ratio of 0.935 and a standard deviation of 0.198). Meanwhile, the shear strength predicted by ML models was then compared with classical mechanics-driven shear models, including two prominent practice codes (i.e., ACI318, EC2) and two previous mechanical models, which indicated that the ML approach is highly reliable and accurate over conventional methods. In addition, a reliability-based design was conducted on two ML models, and their reliability results were compared with those of two code standards. The findings revealed that the ML models demonstrate higher reliability compared to code standards.

Stock Price Trend Analysis and Prediction using Machine Learning

Stock price prediction and trend analysis are crucial aspects of financial markets, presenting complex challenges due to their dynamic nature . This research delves into the application of machine learning algorithms in addressing these challenges, focusing on popular models such as linear regression, Long Short-Term Memory (LSTM), and Convolutional Neural Network (CNN). By leveraging machine learning capabilities, this study aims to enhance predictive accuracy and understand stock market dynamics more comprehensively. The investigation involves a detailed examination and comparison of these algorithms, with linear regression providing a foundational benchmark against advanced techniques like LSTM and CNN, tailored for time series and complex pattern recognition, respectively

A Survey of Google Cloud Platform (GCP): Features, Services, and Applications

This survey paper offers a comprehensive overview of Google Cloud Platform (GCP), examining its extensive range of services, features, and practical applications in various business contexts. GCP is renowned for its diverse offerings in cloud computing, encompassing scalable compute resources, robust storage solutions, advanced machine learning capabilities, and sophisticated data analytics tools. The paper includes a comparative analysis of GCP with other prominent cloud platforms, highlighting its distinctive strengths and strategic positioning in the competitive cloud market. By synthesizing current research and empirical insights, this survey aims to provide a nuanced understanding of GCP's technological advancements, strategic advantages, and implications for enterprises adopting cloud solutions.

Benchmarking the performance and uncertainty of machine learning models in estimating scour depth at sluice outlets

ABSTRACT This study investigates the performance of six machine learning (ML) models – Random Forest (RF), Adaptive Boosting (ADA), CatBoost (CAT), Support Vector Machine (SVM), Lasso Regression (LAS), and Artificial Neural Network (ANN) – against traditional empirical formulas for estimating maximum scour depth after sluice gates. Our findings indicate that ML models generally outperform empirical formulas, with correlation coefficients (CORR) ranging from 0.882 to 0.944 for ML models compared with 0.835–0.847 for empirical methods. Notably, ANN exhibited the highest performance, followed closely by CAT, with a CORR of 0.936. RF, ADA, and SVM performed competitive metrics around 0.928. Variable importance assessments highlighted the dimensionless densimetric Froude number (Fd) as significantly influential, particularly in RF, CAT, and LAS models. Furthermore, SHAP value analysis provided insights into each predictor's impact on model outputs. Uncertainty assessment through Monte Carlo (MC) and Bootstrap (BS) methods, with 1,000 iterations, indicated ML's capability to produce reliable uncertainty maps. ANN leads in performance with higher mean values and lower standard deviations, followed by CAT. MC results trend towards optimistic predictions compared with BS, as reflected in median values and interquartile ranges. This analysis underscores the efficacy of ML models in providing precise and reliable scour depth predictions.

The Expanding Attack Surface: Securing AI and Machine Learning Systems in Security Operations

Cyber threats' increasing magnitude and intricacy require a fundamental change in security operations. Conventional approaches face difficulties in keeping up, which exposes organizations to risks. This paper examines the expanding attack surface: securing AI and machine learning systems in security operations as a remedy. A literature review, informed by the Diffusion of Innovation Theory, investigates how organizations absorb innovations in this study. The results demonstrate notable benefits of AI/ML in security, such as superior identification of threats, improved efficiency through automation, and optimized management of vulnerabilities. Nevertheless, achieving successful execution necessitates meticulous deliberation of obstacles. These tasks encompass guaranteeing data accuracy, preserving the capacity to understand how models work, reducing any potential prejudices in AI/ML models, and resolving security weaknesses in the systems themselves. The paper also discusses ethical considerations and emphasizes the important function of human monitoring. To address these difficulties, the study recommends prioritizing data quality, utilizing explainable AI methods, and developing tactics to mitigate bias. Furthermore, there is a strong emphasis on using a human-in-the-loop strategy to take advantage of humans' expertise and machine-learning capabilities. This study highlights the capacity of artificial intelligence and machine learning to transform security operations completely. By confronting the recognized obstacles, organizations may unleash the genuine potential of these technologies and establish a stronger and more proactive security position in response to constantly changing cyber threats.

Developing a three stage coordinated approach to enhance efficiency and reliability of virtual power plants

A Virtual Power Plant (VPP) is a centralized energy system that manages, and coordinates distributed energy resources, integrating them into a unified entity. While the physical assets may be dispersed across various locations, the VPP integrates them into a virtual unified entity capable of responding to grid demands and market signals. This paper presents a tri-level hierarchical coordinated operational framework of VPP. Firstly, an Improved Pelican Optimization Algorithm (IPOA) is introduced to optimally schedule Distributed Energy Resources (DERs) within the VPP, resulting in a significant reduction in generation costs. Comparative analysis against conventional algorithms such as Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) demonstrates IPOA's superior performance, achieving an average reduction of 8.5% in generation costs across various case studies. The second stage focuses on securing the optimized generation data from rising cyber threats, employing the capabilities of machine learning, preferably, a convolutional autoencoder to learn the normal patterns of the optimized data to detect deviations from the optimized generation data to prevent suboptimal decisions. The model exhibits exceptional performance in detecting manipulated data, with a False Positive Rate (FPR) of 1.92% and a Detection Accuracy (DA) of 98.06%, outperforming traditional detection techniques. Lastly, the paper delves into the dynamic nature of the day ahead market that the VPP participates in. In responding to the grid by selling its optimized generated power via the day-ahead market, the VPP employs the Prophet model, another machine learning technique to forecast the spot market price for the day-ahead to mitigate the adverse effects of price volatility. By utilizing Prophet forecasts, the VPP achieves an average revenue increase of 15.3% compared to scenarios without price prediction, emphasizing the critical role of predictive analytics in optimizing economic gains. This tri-level coordinated approach adopted addresses key challenges in the energy sector, facilitating progress towards achieving universal access to clean and affordable energy.

Innovation Strategy After IPO: How AI Analytics Spurs Innovation After IPO

We examine the role of AI analytics in facilitating innovation in firms that have gone through IPO. Using patent data on over 1,000 publicly traded firms, we find that firms acquiring AI analytics capability post-IPO experience less of a decline in innovation quality compared with similar firms that have not acquired that capability. This effect is greater when only machine learning capabilities are considered. Moreover, we find this sustained rate of innovation is driven principally by the continued development of innovations that combine existing technologies into new ones—a form of innovation that is especially well supported by analytics. By examining three main mechanisms that hampered post-IPO innovation, we find that AI analytics can ameliorate the pressure to meet short-term financial goals and disclosure requirements. However, it has limited effect in addressing managerial incentives. For firms with long product cycles, the disclosure effect is reduced to a greater extent than it is for those with short cycles. Overall, our results show the importance of examining technology as a critical input factor in innovation. We show that the increased deployment of AI analytics may reduce some of the innovative penalties suffered by IPOs and that investors and managers can potentially mitigate post-IPO reductions in innovative output by directing capital acquired in the IPO process to the acquisition of AI analytics capabilities. This paper was accepted by D. J. Wu, information systems. Funding: The authors appreciate the generous financial support from Wharton Dean’s Research Fund and Mack Institute for Innovation Management. Supplemental Material: The data files are available at https://doi.org/10.1287/mnsc.2022.01559 .

How fair is machine learning in credit lending?

AbstractMachine learning models are widely used to decide whether to accept or reject credit loan applications. However, similarly to human‐based decisions, they may discriminate between special groups of applicants, for instance based on age, gender, and race. In this paper, we aim to understand whether machine learning credit lending models are biased in a real case study, that concerns borrowers asking for credits in different regions of the United States. We show how to measure model fairness using different metrics, and we explore the capability of explainable machine learning to add further insights. From a constructive viewpoint, we propose a propensity matching approach that can improve fairness.

Investigating the Impact of Machine Learning in Personalized Education Systems

Personalized education systems leverage robust capabilities of machine learning (ML). Tailoring learning experiences to individual needs preferences and abilities. This survey explores integration of ML in personalized education. It examines methodologies. It highlights applications impacts Future directions are discussed. Key ML techniques include supervised learning for performance prediction and feedback personalization. Unsupervised learning for identifying learning styles. Reinforcement learning.Personalized education systems leverage robust capabilities of machine learning (ML). Tailoring learning experiences to individual needs preferences and abilities. This survey explores integration of ML in personalized education. It examines methodologies. Highlighting applications. Impacts and future directions are discussed. Key ML techniques include supervised learning for performance prediction, feedback personalization unsupervised learning for identifying learning styles, reinforcement learning. Optimizing content sequences. Deep learning for tasks like automated grading. Intelligent tutoring included. Applications such as intelligent tutoring systems. Adaptive learning platforms. Learning analytics. Personalized learning paths and automated assessment systems discussed. The impact of ML-driven personalized education systems is profound. They enhance engagement. Motivation. Academic performance and reduce dropout rates. Challenges such as data privacy algorithmic bias, scalability and resistance to change are also addressed. Future research directions include interdisciplinary collaboration. Advancements in natural language processing are essential. Enhanced personalization algorithms are needed. Long-term studies. Considered crucial. Areas for ongoing work. Key Words: Machine learning, Personalized education, Adaptive learning, Intelligent tutoring systems, Predictive analytics, Recommendation systems, Ethical considerations, Student engagement, Educational technology

Stability Assessment of an Ore Mill Electric Drive Using Machine Learning

The relevance of the study is due to the need to improve electric drive systems operated in harsh conditions. The goal of the study is to create a model for assessing the state of stability of the electric drive of an ore mill using machine learning capabilities, which will provide high performance and the ability to work consistently in different systems. Various sustainability assessment models have been developed based on 6 machine learning algorithms. The study and comparison of models built using artificial neural networks (ANN) of different architectures was carried out using various learning methods. The expediency of using the Tree and ANN algorithms to develop a model for assessing electric drive stability is substantiated. The novelty of the results obtained lies in the fact that the model has high accuracy, high speed, and the ability to detect instability in uncertain operating modes of the electric motor of an electric drive, as well as the possibility of coordinated operation with various systems. The practical value is that the model allows, at an intellectual level, to provide effective control and fault diagnosis of complex electric drive systems, which cannot be achieved using the known methods. Doi: 10.28991/HIJ-2024-05-02-01 Full Text: PDF

National implementation of an AI-based virtual dietitian for patients with cancer.

1500 Background: Nutritional status is an established key driver of cancer outcomes and patient experience. However, the workforce of oncology dietitians is unable to meet the needs of the population, and 80% of patients seek but do not receive nutrition support. Advances in artificial intelligence (AI) and machine learning (ML) capabilities afford the opportunity to expand reach and rapidly hone in on clinical and contextual needs to deliver evidence-based nutritional guidance. Methods: To address this unmet need, an AI-based nutrition expert-platform was developed. Interventions were curated by oncology-nutrition experts based on peer-reviewed literature and clinical guidelines to facilitate self-management of cancer side effects and optimize nutrition. A growing database of over 114,000 evidence-based interventions powers Ina, a text-message based virtual nutrition assistant. Interventions are continually enhanced and refined using AI and ML applied to an extensive learning dataset. The tool was implemented nationally in partnership with 25 cancer advocacy organizations. Data on demographics, patient-reported outcomes, and utilization were systematically collected. Results: Between July 2019 and August 2023, 3,310 users from all 50 states registered for the virtual nutrition platform. Based on self-report, users were 73% female; mean age 57 years (18-91); most common cancer types were GU (22%), Breast (21%), Gyn (19%), GI (14%), and Lung (12%), representing nearly 90% of users. Users were medically complex, with 50% reporting Stage 3-4 disease, 64% with metastases, and 50% with 2+ chronic conditions. Nutritional challenges were highly prevalent: 58% were overweight/obese by BMI, 83% reported barriers to good nutrition, and 42% had food allergies/intolerances. Levels of engagement were high: 68% of users texted queries to Ina, 79% completed surveys, and median user retention duration was 8.8 months. Overall, 94% of users reported being satisfied with the platform and 98% found the guidance to be helpful. In an evaluation of outcomes, 84% reported actively using the nutrition advice to guide their diets, 47% used the recommended recipes, and 82% felt the program improved their quality of life. In this cohort with high symptom burden, 88% reported Ina helped them manage their symptoms and 34% noted improved symptom control during participation. Conclusions: Implementation of an evidence-based AI virtual dietitian is feasible, and is reported by patients to be beneficial on diet, quality of life, and symptom control. Ongoing evaluations are assessing impact on other outcomes.

Artificial Intelligence in Central Banking

Abstract The paper uses qualitative research to investigate the potential uses of artificial intelligence in the field of central banking. The analysis shows that monetary policy, prudential supervision and the oversight of payments are the areas where the use artificial intelligence is most likely to bring benefits. Monetary policy calibration involves working with long time series of data for various parameters and making the necessary analysis and forecasts, an activity in which artificial neural networks may prove useful. Bank supervision can benefit from the use natural language processing algorithms that can read documents and extract the relevant information. Such algorithms can read all of the required documents (not just those that the supervisor selected) and return all of the sentences that contain a certain predefined expression. In the field of the oversight of payments, the capabilities of machine learning to identify new patterns or anomalies in the data that could indicate fraud or money laundering will boost the efforts to combat them. In terms of challenges associated with the use of artificial intelligence in central banking, perhaps the two biggest challenges are that some of the models do not allow for a reasonable level of explainability of the algorithm(s) through which they arrive at the result (especially relevant for bank supervision) and data availability. With respect to the latter, although the issue of quantity of data can be dismissed as a shortcoming given the huge amounts of data available, the issue of data quality seems to be more pronounced, as deficiencies such as data measured incompletely or incorrectly, scarcity and regulatory barriers that impede data sharing may be difficult to surpass.

Predicting the pulse of the Amazon: Machine learning insights into deforestation dynamics

This study aims to analyze deforestation in the Brazilian Amazon from 1999 to 2020 using machine learning techniques to assess 16 critical factors. Our approach leverages the capabilities of machine learning, particularly Random Forest, which proved to be the most accurate model in terms of determination coefficient, mean squared error, and mean absolute error. The analysis revealed that the harvested area of permanent crops is the most influential variable in predicting deforestation, followed by the area of temporary crops. Furthermore, our findings indicate a significant inverse relationship between public spending and deforestation rates. These results contribute to understanding deforestation dynamics and offer potential strategies for improving conservation efforts.