Combining Machine Learning Research Articles

Combining machine learning and molecular simulation to explore MOF materials that contribute to CF4/N2 separation

High-throughput molecular simulations and machine learning algorithms have been widely used to identify promising metal–organic frameworks for gas separation. However, most studies are limited to screening high-performance materials from existing databases, which fails to fully utilize the predictive function of machine learning. This paper combines genetic algorithms and high-throughput screening to deeply mine anion-pillared MOF (APMOF) performance feature relationships and predict high-performance materials that are not in the database. Considering the actual industrial conditions (CF4/N2 = 10/90), we chose the ratio of CF4:N2 = 1:9 to simulate the adsorption separation of gas mixtures of MOF materials at a temperature of 298 K and a pressure of 1 bar. First, the CF4/N2 separation properties of MOFs in the APMOF library were obtained based on molecular simulations. Then, the filtered data were coded according to the method of “building block classification − structural interval categorization”. Then, the machine learning algorithm was used for model training to obtain a high-precision model. Finally, the tangent adaptive genetic algorithm was used to recombine the genes of the materials, and the new MOF materials were successfully reverse-engineered. The study found that the pore-limit diameter of APMOFs is most conducive to the separation of CF4/N2 by MOFs when the pore-limit diameter of APMOFs is within two times the molecular dynamics diameter of CF4. 134 MOF materials were predicted to have CF4/N2 selectivities exceeding 46.30. The use of organic ligands such as 4,4′-bipyridyl or 1,4-bis(4-pyridyl)benzene (bpb) increases the likelihood of these materials being high-performance for CF4/N2 separation. The combination of computational screening methods and machine learning can expedite the design and development of new high-performance MOFs. This approach can also provide guidance for the synthetic direction and pattern of materials, which can assist experimentalists in their synthesis.

Highly stable silicon oxycarbide all-solid-state batteries enabled by machined learning accelerated screening of oxides and sulfides electrolytes

Traditional trial–error approach severely limits and restricts rapid development of high-performance anode and electrolytes materials, searching huge parameters space of various anode-solid electrolyte interfaces in an effective and efficient way is the key issue. Here, a novel computational strategy combining machine learning and first-principles is proposed to achieve efficient high-throughput screening of oxides and sulfides electrolytes for highly stable silicon oxycarbide all-solid-state batteries. First-principles calculations demonstrate significant compact of material type and elemental doping on interfacial compatibility between silicon oxycarbide and various electrolytes. By proposing several novel descriptors including interfacial adhesion and formation energies of frozen system with low computation cost, the amounts of demanded trainings data are significantly reduced. Gradient-boosted regression tree model shows low mean absolute errors of 0.09 and high R2 value of 0.99 for the prediction of interface formation energy, demonstrating ultrahigh accuracy and reliability of the algorithm. The present work discovers a series of uninvestigated stable anode-solid electrolytes interfacial couples for further experimental preparation.

MXene-Based Skin-Like Hydrogel Sensor and Machine Learning-Assisted Handwriting Recognition.

Conductive hydrogels are widely used in flexible sensors owing to their adjustable structure, good conductivity, and flexibility. The performance of excellent mechanical properties, high sensitivity, and elastic modulus compatible with human tissues is of great interest in the field of flexible sensors. In this paper, the functional groups of trisodium citrate dihydrate (SC) and MXene form multiple hydrogen bonds in the polymer network to prepare a hydrogel with mechanical properties (Young's modulus (23.5-92 kPa) of similar human tissue (0-100 kPa)), sensitivity (stretched GF is 4.41 and compressed S1 is 5.15 MPa-1), and durability (1000 cycles). The hydrogel is able to sensitively detect deformations caused by strain and stress and can be used in flexible sensors to detect human movement in real time such as fingers, wrists, and walking. In addition, the combination of matrix sensing and machine learning was successfully used for handwriting recognition with an accuracy of 0.9744. The combination of machine learning and flexible sensors shows great potential in areas such as healthcare, information security, and smart homes.

Computational Test for Conditional Independence

Conditional Independence (CI) testing is fundamental in statistical analysis. For example, CI testing helps validate causal graphs or longitudinal data analysis with repeated measures in causal inference. CI testing is difficult, especially when testing involves categorical variables conditioned on a mixture of continuous and categorical variables. Current parametric and non-parametric testing methods are designed for continuous variables and can quickly fall short in the categorical case. This paper presents a computational approach for CI testing suited for categorical data types, which we call computational conditional independence (CCI) testing. The test procedure is based on permutation and combines machine learning prediction algorithms and Monte Carlo cross-validation. We evaluated the approach through simulation studies and assessed the performance against alternative methods: the generalized covariance measure test, the kernel conditional independence test, and testing with multinomial regression. We find that the computational approach to testing has utility over the alternative methods, achieving better control over type I error rates. We hope this work can expand the toolkit for CI testing for practitioners and researchers.

Assessment of Published Papers on the Use of Machine Learning in Diagnosis and Treatment of Mastitis

The present study is an evaluation of published papers on machine learning as employed in mastitis research. The aim of this study was the quantitative evaluation of the scientific content and the bibliometric details of these papers. In total, 69 papers were found to combine machine learning in mastitis research and were considered in detail. There was a progressive yearly increase in published papers, which originated from 23 countries (mostly from China or the United States of America). Most original articles (n = 59) referred to work involving cattle, relevant to mastitis in individual animals. Most articles described work related to the development and diagnosis of the infection. Fewer articles described work on the antibiotic resistance of pathogens isolated from cases of mastitis and on the treatment of the infection. In most studies (98.5% of published papers), supervised machine learning models were employed. Most frequently, decision trees and support vector machines were employed in the studies described. ‘Machine learning’ and ‘mastitis’ were the most frequently used keywords. The papers were published in 39 journals, with most frequent publications in Computers and Electronics in Agriculture and Journal of Dairy Science. The median number of cited references in the papers was 39 (interquartile range: 31). There were 435 co-authors in the papers (mean: 6.2 per paper, median: 5, min.–max.: 1–93) and 356 individual authors. The median number of citations received by the papers was 4 (min.–max.: 0–70). Most papers (72.5%) were published in open-access mode. This study summarized the characteristics of papers on mastitis and artificial intelligence. Future studies could explore using these methodologies at farm level, and extending them to other animal species, while unsupervised learning techniques might also prove to be useful.

Fast dynamic prediction of consequences of heavy gas leakage accidents based on machine learning

The field of emergency risk management in chemical parks has been characterized by a lack of fast, precise and dynamic prediction methods. The application of computational fluid dynamics (CFD) models, which offer the potential for dynamic and precise prediction, has been hindered by high computational costs. Therefore, taking liquid benzene as a case study, this paper combined machine learning (ML) algorithms with a CFD-based precise prediction model, to develop an ML model for fast dynamic prediction of heavy gas leakage consequences in chemical parks. Employing the CFD data as the input, the prediction models were developed using ML algorithms, refined with Bayesian optimization for parameter tuning. This study utilized PHOENICS software to establish a dynamic prediction model for the diffusion of liquid benzene leakage, validated by Burro nine experiment data. Comparative analyses of models based on five ML algorithms were conducted to evaluate the reliability of their predictions using both CFD standard and noisy data. The results indicated that temperature had the most significant effect on the consequences of the leakage accidents among four key factors (wind speed, temperature, leakage aperture and atmospheric stability), followed by wind speed. These factors served as input variables for ML model training. Among the models evaluated, the eXtreme Gradient Boosting (XGBoost) model showed superior performance, irrespective of the presence of noise in the data. An XGBoost-based fast prediction model was ultimately developed for predicting the consequences of liquid benzene leakage. A case analysis was conducted to validate the feasibility of the model prediction. The relative errors between the predicted and actual values of the model for acute exposure guideline level-1 (AEGL-1), AEGL-2, and AEGL-3 distances were 2.70%, 2.58%, and 0.23%, respectively. Furthermore, the XGBoost model completed the prediction in only 0.218 s, a stark contrast to the hours necessitated by the CFD model, thus offering substantial computational time savings while maintaining high accuracy levels. This paper introduces an ML model for fast dynamic prediction of heavy gas leakage, enabling chemical parks to make more timely and accurate decisions in emergency risk management.

Synergistic Biophysics and Machine Learning Modeling to Rapidly Predict Cardiac Growth Probability.

Computational models that can predict growth and remodeling of the heart could have important clinical applications. However, the time it takes to calibrate and run current models while considering data uncertainty and variability makes them impractical for routine clinical use. This study aims to address this need by creating a computational framework to efficiently predict cardiac growth probability. We utilized a biophysics model to rapidly simulate cardiac growth following mitral valve regurgitation (MVR). Here we developed a two-tiered Bayesian History Matching approach augmented with Gaussian process emulators for efficient calibration of model parameters to align with growth outcomes within a 95% confidence interval. We first generated a synthetic data set to assess the accuracy of our framework, and the effect of changes in data uncertainty on growth predictions. We then calibrated our model to match baseline and chronic canine MVR data and used an independent data set to successfully validate the ability of our calibrated model to accurately predict cardiac growth probability. The combined biophysics and machine learning modeling framework we proposed in this study can be easily translated to predict patient-specific cardiac growth.

Combination of ESG scores and prediction-based returns using long short-term memory neural networks to generate responsible portfolios

ABSTRACT The significance of the environmental, social, and governance (ESG) factors has risen substantially among investors in recent years. Similarly, machine learning techniques have allowed for improvements in the prediction of stock prices. Our research combines ESG factors and prediction-based returns using recurrent neural networks to create profitable-sustainable portfolios that can consistently beat the market index in return and ESG scores. Our analysis focuses on the components of the EURO STOXX 50Ⓡ Index during the year 2021 and the first half of 2022, allowing us to analyze two different market scenarios, continuous growth and bear market, respectively. This paper provides empirical evidence that combining machine learning with ESG scores and its application in portfolio optimization can achieve higher returns and higher ESG performance depending on the macroeconomic context and presents the trade-off between the Sharpe Ratio and the ESG score of the optimized portfolios for different scenarios.

An Efficient Customer Churn Prediction Technique Using Combined Machine Learning in Commercial Banks

An Efficient Customer Churn Prediction Technique Using Combined Machine Learning in Commercial Banks

High frequency monitoring of credit creation: A new tool for central banks in emerging market economies

This study utilizes weekly datasets on loan growth in Colombia to develop a daily indicator of credit expansion using a two-step machine learning approach. Initially, employing Random Forests (RF), missing data in the raw credit indicator is filled using high frequency indicators like spreads, interest rates, and stock market returns. Subsequently, Quantile Random Forest identifies periods of excessive credit creation, particularly focusing on growth quantiles above 95 %, indicative of potential financial instability. Unlike previous studies, this research combines machine learning with mixed frequency analysis to create a versatile early warning instrument for identifying instances of excessive credit growth in emerging market economies. This methodology, with its ability to handle nonlinear relationships and accommodate diverse scenarios, offers significant value to central bankers and macroprudential authorities in safeguarding financial stability.

TriStack enables accurate identification of antimicrobial and anti-inflammatory peptides by combining machine learning and deep learning approaches

The identification of antimicrobial peptides (AMPs) and anti-inflammatory peptides (AIPs) is crucial for drug design and disease treatment. However, it remains a computational challenge to accurately identify these peptides due to insufficient information encoding the peptide sequences. In this study, we propose TriStack, a powerful and interpretable model for accurate identification of AMPs and AIPs by stacking a machine learning-based module using a multi-layer residual network. It first extracts three types of function-related features from peptide sequences to comprehensively characterize the composition, distribution, and physicochemical properties of residues. Furthermore, these features are fused and fed into a two-module stacked model. The first module provides the preliminary predictions based on three machine learning methods, while the second module refines these predictions further via a multi-layer residual network. After training and testing, TriStack outperforms all the compared leading methods for both AMPs and AIPs predictions. TriStack is expected to contribute to antimicrobial and anti-inflammatory drug based on peptide sequences.

Monitoring soil nutrients using machine learning based on UAV hyperspectral remote sensing

ABSTRACT Unmanned aerial vehicles (UAV) are rapidly evolving experimental platforms that play an important role in remote sensing. In this study, we investigated a machine learning method for monitoring soil nutrient content using UAV hyperspectral remote sensing. We employed machine learning techniques for feature extraction and soil hyperspectral information modelling. In contrast to traditional mathematical transformation methods, we adopted a combination of random forest and differential evolution algorithms to rank the weights of individual hyperspectral data, thereby obtaining a series of spectral feature subsets for soil organic matter, total nitrogen and available phosphorus and potassium. Furthermore, the analytic hierarchy process was used for weight analysis, and the characteristic bands of the four soil nutrients were successfully extracted. Next, a quantitative inversion model based on a back-propagation (BP) neural network was established to estimate soil nutrient content, with determination coefficients higher than 0.7 and 0.6 for the modelling and verification sets, respectively. The relative percent difference values were greater than 2, among which the highest was for available potassium, with determination coefficients of 0.95 and 0.84 for the modelling and verification sets, respectively. In addition, visualization distribution maps of soil nutrients were obtained by combining the BP model and original reflectance hyperspectral images, and the comparisons of content histograms showed a relatively consistent distribution between the sampling and inversion points. The results verified the effectiveness of the combined machine learning method for large-scale and high-precision monitoring and visualization of soil nutrient contents.

Detection of Disease in Calves using Artificial Intelligence

Background: Livestock farming is experiencing a digital transformation and is becoming more information-driven. However, this type of data is often kept in separate storage towers, making it incapable of practicing its potential to boost animal welfare. Lumpy Skin Disease (LSD) is a serious threat to the health of cattle worldwide and has caused financial problems for many cattle farming enterprises. It has been shown that combining machine learning (ML) and artificial intelligence (AI) with biosensor data and conventional visual inspections can improve the identification and diagnosis of this serious condition. Methods: This study presents an extremely precise livestock farming framework that combines data streams from a wide array of disciplines of dairy cattle to see if ever wider/vast data sources enhance the overall projections for diseases and if using the more complicated prediction models can reimburse for less diverse data to some extent. Using images from the farming landscape, this study highlights the utility of convolutional neural networks (CNNs) in the identification of the Lumpy Skin Disease Virus (LSDV) in animals. Result: An analysis of the relative weight given to individual factors in predicting accuracy shows that disease in dairy cattle results from the intricate interactions between many life domains/parameters, such as housing, nutrition and climate; that prediction performance is enhanced by incorporating a wider range of data sources; and that current information can be repurposed to produce useful information for vaccine development. The study highlights the potential of data-driven dairy interventions, focusing on artificial intelligence for disease prediction in cattle, to improve animal welfare and reduce the risk of disease. A Convolutional Neural Network (CNN) based model effectively classified skin conditions with an overall accuracy of 73.89% after 27 training epochs. This study demonstrates CNN’s useful applications in the field of veterinary medicine by highlighting its potential for early detection of Lumpy Skin Disease (LSD).

Framework for Mitigating Phishing E-mail in the Kenyan Banking Industry Using Artificial Intelligence (AI)

Purpose: Phishing is a significant cybercrime threat that affects individuals and organizations globally, including the banking industry in Kenya. The sophistication of phishing attacks continues to increase, and it is increasingly challenging traditional security measures to mitigate these threats. The purpose of this thesis is to build a framework for mitigating phishing e-mail attacks in the banking industry in Kenya using artificial intelligence. Phishing emails are among the most common techniques of cyber-attacks utilized by assailants to gain unauthorized access to sensitive information such as financial details, personal information, and login credentials. These attacks can have devastating effects on the victims, leading to financial loss, reputation damage, and even identity theft. Methodology: The framework development consists of four main stages: data collection, data preprocessing, model training, and deployment. In the data collection stage, a dataset of phishing and non-phishing emails is gathered from various sources such as public databases, dark web forums, and bank employees mail. In the data preprocessing stage, the collected data is cleaned, preprocessed, and labeled. In the model training stage, machine learning algorithms and NLP techniques is used to develop a robust phishing and non-phishing emails detection model. In the deployment stage, the model is integrated into the bank's email system to detect and block phishing emails in real-time. The framework is then evaluated using a dataset of phishing and non-phishing e-mails collected from the banking industry in Kenya. Various metrics such as accuracy, precision, recall, and F1-score are used to evaluate the framework. The framework is able to detect new phishing e-mails that were not previously included in the dataset, demonstrating its ability to adapt to new threats. Findings: The framework is based on a hybrid approach that combines machine learning algorithms, natural language processing (NLP) techniques, and human expertise that identify and prevent phishing emails from reaching their targets. The four main components of this framework include e-mail filtering, feature extraction, classification, and response. The e-mail filtering component uses several algorithms to identify and filter suspicious e-mails. The feature extraction component analyzes the content of the e-mail and extracts relevant features to help classify the e-mail as either legitimate or phishing. The classification component uses machine-learning algorithms to classify the e-mail as either legitimate or phishing. Finally, the response component takes appropriate action based on the classification results. Unique Contribution to Theory, Practice and Policy: The framework provides an effective way to identify and mitigate phishing e-mail attacks, reducing the risk of data breaches and financial losses.

Setting the morphologic quality limits enabling accurate classification of charred archaeological grape seeds

This study investigates the morphological changes in grape pips resulting from various charring conditions. Employing high-resolution scanning combined with morphometric measurements for morphological analysis, we aimed to understand the effects of charring on grape pips. Our morphometric analysis demonstrated significant alterations in seed shape above 250 °C. The length–width ratio and the occurrence of cracks notably changed, providing a basis for assessing charring conditions. In addition, applying a machine learning classification method, we determined that accurate classification of grape varieties by the morphometric analysis method is feasible for seeds charred at up to 250 °C and 8 h. Integrating the morphometric changes and temperature ranges suitable for classification, we developed a sorting model for archaeological seeds. By projecting length–width ratios onto a curve calculated from controlled conditions, we estimated charring temperatures. Approximately 50% of archaeological seeds deviated from the model, indicating drastic charring conditions. This sorting model facilitates a stringent selection of seeds fit for classification, enhancing the accuracy of our machine learning-based methodology. In conclusion, combining machine learning with morphometric sorting enables the identification of charred grape seeds suitable for identification by the morphometric method. This comprehensive approach provides a valuable tool for future research for the identification of charred grape seeds found in archaeological contexts, enhancing our understanding of ancient viticulture practices and grape cultivation.

A hybrid model for the detection of retinal disorders using artificial intelligence techniques

The prevalence of vision impairment is increasing at an alarming rate. The goal of the study was to create an automated method that uses optical coherence tomography (OCT) to classify retinal disorders into four categories: choroidal neovascularization, diabetic macular edema, drusen, and normal cases. This study proposed a new framework that combines machine learning and deep learning-based techniques. The utilized classifiers were support vector machine (SVM), K-nearest neighbor (K-NN), decision tree (DT), and ensemble model (EM). A feature extractor, the InceptionV3 convolutional neural network, was also employed. The performance of the models was evaluated against nine criteria using a dataset of 18000 OCT images. For the SVM, K-NN, DT, and EM classifiers, the analysis exhibited state-of-the-art performance, with classification accuracies of 99.43%, 99.54%, 97.98%, and 99.31%, respectively. A promising methodology has been introduced for the automatic identification and classification of retinal disorders, leading to reduced human error and saved time.

Exploration of Machine Learning and Deep Learning Approaches in Medical Domain

Through the combination of machine learning (ML) and deep learning (DL) approaches, substantial progress has been made in the field of medical picture categorization, which is an essential component in the field of medical diagnostics. Within the context of medical picture categorization, this paper provides an in-depth examination of the development, methodology, and applications of machine learning and deep learning. By making use of handmade features, traditional machine learning techniques, such as support vector machines and decision trees, have laid the groundwork for early achievements in the field. On the other hand, the introduction of deep learning, and more specifically convolutional neural networks (CNNs), has brought about a revolution in the industry by making it possible to automatically extract features and obtaining greater performance. This article takes a look at a number of different deep learning architectures, including ResNet, VGG, and Inception, and highlights the contributions that these designs have made to tasks such as illness categorization, organ segmentation, and tumor identification. In addition to this, it discusses alternative solutions such as data augmentation, transfer learning, and model optimization after addressing problems such as the lack of data, the interpretability of the data, and the demands placed on the computing resources. In addition, the evaluation takes into account the ethical concerns, as well as the need for rigorous validation in order to guarantee clinical application. This study highlights the revolutionary influence that machine learning and deep learning have had on medical imaging by conducting a comparative analysis of current research. It also highlights the ongoing need for innovation and cooperation across disciplines in order to improve diagnostic accuracy and patient outcomes.

Integrating Shapley Values into Machine Learning Techniques for Enhanced Predictions of Hospital Admissions

(1) Background: Predictive modeling is becoming increasingly relevant in healthcare, aiding in clinical decision making and improving patient outcomes. However, many of the most potent predictive models, such as deep learning algorithms, are inherently opaque, and their decisions are challenging to interpret. This study addresses this challenge by employing Shapley Additive Explanations (SHAP) to facilitate model interpretability while maintaining prediction accuracy. (2) Methods: We utilized Gradient Boosting Machines (GBMs) to predict patient outcomes in an emergency department setting, with a focus on model transparency to ensure actionable insights. (3) Results: Our analysis identifies “Acuity”, “Hours”, and “Age” as critical predictive features. We provide a detailed exploration of their intricate interactions and effects on the model’s predictions. The SHAP summary plots highlight that “Acuity” has the highest impact on predictions, followed by “Hours” and “Age”. Dependence plots further reveal that higher acuity levels and longer hours are associated with poorer patient outcomes, while age shows a non-linear relationship with outcomes. Additionally, SHAP interaction values uncover that the interaction between “Acuity” and “Hours” significantly influences predictions. (4) Conclusions: We employed force plots for individual-level interpretation, aligning with the current shift toward personalized medicine. This research highlights the potential of combining machine learning’s predictive power with interpretability, providing a promising route concerning a data-driven, evidence-based healthcare future.

The underlying signals of efficiency in European universities: a combined efficiency and machine learning approach

ABSTRACT This study explores the critical determinants of efficiency in European Higher Education Institutions (HEIs). The contemporary landscape presents a wide range of challenges and opportunities, necessitating unprecedented levels of efficiency. Our primary objective is to unravel the fundamental elements that drive the efficiency of European HEIs. To achieve this, we employ a novel methodological approach, integrating Data Envelopment Analysis (DEA) with a Machine Learning technique, specifically Feature Selection. This combination sheds new light on areas previously less explored in DEA, including the intricate interplay among variables. Our empirical investigation contributes to the academic discourse by presenting a comprehensive, multi-country analysis from a multifaceted viewpoint. We assess the efficiency levels of European universities, highlighting the influence of factors such as student fee funding, Ph.D. program intensity, and international mobility on achieving high efficiency scores. The findings of this study have some implications for policy-making, suggesting strategies to enhance the efficiency levels of HEIs.

Leveraging Quantitative Systems Pharmacology and Artificial Intelligence to advance treatment of Chronic Kidney Disease Mineral Bone Disorder.

Chronic kidney disease mineral bone disorder (CKD-MBD) is a complex clinical syndrome responsible for the accelerated cardiovascular mortality seen in individuals afflicted with CKD. Current approaches to therapy have failed to improve clinical outcomes adequately, likely due to targeting surrogate biochemical parameters as articulated by the guideline developer, KDIGO (Kidney Disease: Improving Global Outcomes). We hypothesized that using a Systems Biology Approach combining machine learning with mathematical modeling, we could test a novel approach to therapy targeting the abnormal movement of mineral out of bone and into soft tissue that is characteristic of CKD-MBD. The mathematical model describes the movement of calcium and phosphate between body compartments in response to standard therapeutic agents. The machine learning technique we applied is Reinforcement Learning (RL). We compared calcium, phosphate, PTH, and mineral movement out of bone and into soft tissue under four scenarios: standard approach (KDIGO), achievement of KDIGO guidelines using RL (RLKDIGO), targeting abnormal mineral flux (RLFLUX), and combining achievement of KDIGO guidelines with minimization of abnormal mineral flux (RLKDIGOFLUX). We demonstrate through simulations that explicitly targeting abnormal mineral flux significantly decreases abnormal mineral movement compared to standard approach while achieving acceptable biochemical outcomes. These investigations highlight the limitations of current therapeutic targets, primarily secondary hyperparathyroidism, and emphasize the central role of deranged phosphate homeostasis in the genesis of the CKD-MBD syndrome.