Machine Learning Classification Algorithms Research Articles

Quantum data encoding: a comparative analysis of classical-to-quantum mapping techniques and their impact on machine learning accuracy

This study explores the integration of quantum data embedding techniques into classical machine learning (ML) algorithms; to assess performance enhancements and computational implications across a spectrum of models. We explored various classical-to-quantum mapping methods; ranging from basis encoding and angle encoding to amplitude encoding; for encoding classical data. We conducted an extensive empirical study encompassing popular ML algorithms, including Logistic Regression, K-Nearest Neighbors, Support Vector Machines, and ensemble methods like Random Forest, LightGBM, AdaBoost, and CatBoost. Our findings reveal that quantum data embedding contributes to improved classification accuracy and F1 scores, particularly notable in models that inherently benefit from enhanced feature representation. We observed nuanced effects on running time, with low-complexity models exhibiting moderate increases and more computationally intensive models experiencing discernible changes. Notably, ensemble methods demonstrated a favorable balance between performance gains and computational overhead.This study underscores the potential of quantum data embedding to enhance classical ML models and emphasizes the importance of weighing performance improvements against computational costs. Future research may involve refining quantum encoding processes to optimize computational efficiency and explore scalability for real-world applications. Our work contributes to the growing body of knowledge on the intersection of quantum computing and classical machine learning, offering insights for researchers and practitioners seeking to harness the advantages of quantum-inspired techniques in practical scenarios.

Design of ultra-broadband long-wave to ultra-long-wave infrared absorber based on machine learning

Abstract long-wave infrared (LWIR) and ultra-long-wave infrared (ULWIR) radiation find extensive utility in atmospheric monitoring, night reconnaissance, deep space exploration, and various other fields. However, achieving ultra-wideband absorption within these spectral ranges has persistently posed a significant challenge for researchers. Metasurfaces have attracted great interest due to their ability to manipulate electromagnetic (EM) waves with unprecedented precision and efficiency. In this paper, a metasurface absorber with a simple structure is proposed. Combined with machine learning (ML) of Random Forest algorithm for design optimization, the absorber achieves ultra-wideband perfect absorption of 42.54 μm within the LWIR-ULWIR band. Furthermore, the stability and accuracy of the Random Forest algorithm in absorber design are also evaluated and compared with other classical ML algorithms. The successful realization of this work can offer the advancement of thermal imaging, sensing, and communication systems operating within the LWIR and ULWIR spectra.

PV Module Soiling Detection Using Visible Spectrum Imaging and Machine Learning

During the last decades photovoltaic solar energy has continuously increased its share in the electricity mix and has already surpassed 5% globally. Even though photovoltaic (PV) installations are considered to require very little maintenance, their efficient exploitation relies on accounting for certain environmental factors that affect energy generation. One of these factors is the soiling of the PV surface, which could be observed in different forms, such as dust and bird droppings. In this study, visible spectrum data and machine learning algorithms were used for the identification of soiling. A methodology for preprocessing the images is proposed, which puts focus on any soiling of the PV surface. The performance of six classification machine learning algorithms is evaluated and compared—convolutional neural network (CNN), support vector machine (SVM), random forest (RF), k-nearest neighbor (kNN), naïve-Bayes, and decision tree. During the training and validation phase, RF proved to be the best-performing model with an F1 score of 0.935, closely followed by SVM, CNN, and kNN. However, during the testing phase, the trained CNN achieved the highest performance, reaching F1 = 0.913. SVM closely followed it with a score of 0.895, while the other two models returned worse results. Some results from the application of the optimal model after specific weather events are also presented in this study. They confirmed once again that the trained convolutional neural network can be successfully used to evaluate the soiling state of photovoltaic surfaces.

Machine Learning Techniques Towards Accurate Emotion Classification from EEG Signals

This article delves into using machine learning algorithms for emotion classification via EEG brain signals. The goal is to discover an accurate model beyond traditional methods, necessitating AI for classifying emotional EEG signals. This study, motivated by the complex link between emotions and neural activity, employs Random Forest, Support Vector Machines, and K-Nearest Neighbors. Notably, Random Forest achieves 99% accuracy, SVM 98%, and KNN 94%. These impressive results, backed by performance metrics like confusion matrices, reveal each model’s effectiveness in emotion classification. The dataset, rich in varied emotional stimuli and EEG placements, provides a robust foundation for detailed analysis. This research underscores significant applications in affective computing and mental health, offering a promising path to understanding the intricate relationship between EEG signals and human emotions.

Tensile strength prediction in monel 400 weldments using classification and regression algorithms in machine learning

Abstract Selecting the most effective welding settings impacts mechanical strength and weld quality, with parameters like current, voltage, and speed playing pivotal roles. The mechanical part encompasses material properties, welding process variables, and structural integrity, all contributing to the overall weld quality and strength. By integrating these mechanical factors with predictive modelling, a comprehensive understanding of weld performance can be achieved, enabling optimized welding settings and enhanced weld quality assurance. This study assesses and compares machine learning algorithms such as a random tree, random forest, and C4.5 to determine their predictive capability regarding the tensile strength in Monel 400 Weldments. By utilizing a dataset comprising 32 instances with attributes like Current, Voltage, and Speed, models were developed and assessed using K-Fold cross-validation. Among these algorithms, the random tree models emerge as the most proficient in accurately predicting the tensile strength for Monel 400 Weldments through classification ML techniques. Similarly, regression algorithms have been deployed to assess the dataset by varying the train-test split ratio and gradient boosting, which exhibited superior performance with a higher R2 value of 0.99. Both random tree and Gradient boosting algorithms have commonly been recommended, with current being the most influential factor affecting tensile strength.

Classification of Garlic Varieties with Fluorescent Spectroscopy Using Machine Learning

Machine learning techniques can produce fast, accurate and objective results in the analysis of agricultural products. These artificial intelligence-based systems are frequently encountered in studies on agriculture in the literature. This study reveals the usability of machine learning algorithms in classification of garlic cultivars using fluorescent spectroscopic data. For this, six types of garlic were used: Razgradski-11, Razgradski-12, Razgradski-115, Plovdivski-120, Yambolski-99 and Topolovgradski. In the first stage, the parsing analysis made from the fluorescent spectroscopic data of the garlics was carried out with seven different machine learning. The classification results of these seven types of machine learning algorithms were obtained. In the second stage, the classification results were obtained by adjusting the hyperparameters of each Machine Learning (ML) algorithm in order to control the improvability of the classification accuracy rates. Finally, performance metrics such as Specificity, precision, MCC, F1-Score of the classification processes obtained in the two stages were compared. In general, it was observed that the classification performances increased with the hyperparameter adjustment performed in the second stage. In this study, classification results with ML showed that fluorescent spectroscopy data of garlic strongly represented garlic species and provided high performance classification accuracy of 99.93% with Neural Network (NN), one of the machine learning methods using these data.

Deep learning models for the prediction of acute postoperative pain in PACU for video‐assisted thoracoscopic surgery

BackgroundPostoperative pain is a prevalent symptom experienced by patients undergoing surgical procedures. This study aims to develop deep learning algorithms for predicting acute postoperative pain using both essential patient details and real-time vital sign data during surgery.MethodsThrough a retrospective observational approach, we utilized Graph Attention Networks (GAT) and graph Transformer Networks (GTN) deep learning algorithms to construct the DoseFormer model while incorporating an attention mechanism. This model employed patient information and intraoperative vital signs obtained during Video-assisted thoracoscopic surgery (VATS) surgery to anticipate postoperative pain. By categorizing the static and dynamic data, the DoseFormer model performed binary classification to predict the likelihood of postoperative acute pain.ResultsA total of 1758 patients were initially included, with 1552 patients after data cleaning. These patients were then divided into training set (n = 931) and testing set (n = 621). In the testing set, the DoseFormer model exhibited significantly higher AUROC (0.98) compared to classical machine learning algorithms. Furthermore, the DoseFormer model displayed a significantly higher F1 value (0.85) in comparison to other classical machine learning algorithms. Notably, the attending anesthesiologists' F1 values (attending: 0.49, fellow: 0.43, Resident: 0.16) were significantly lower than those of the DoseFormer model in predicting acute postoperative pain.ConclusionsDeep learning model can predict postoperative acute pain events based on patients' basic information and intraoperative vital signs.

Advancing source reactor-type discrimination using machine learning techniques and SFCOMPO-2.0 experimental database

In recent years, nuclear forensic analysis has become crucial due to the growing global threat of nuclear terrorism and smuggling. Since 2005, extensive research has been conducted on identifying the origin of spent nuclear fuel, focusing on the source reactor-type discrimination, 235U enrichment of the fresh fuel, and the fuel exposure in the reactor (known as burnup). However, the majority of research relies on computed databases, which may lead to tracing discrepancies compared with actual situations. The present study employs the isotopic measurements from the experimental SFCOMPO-2.0 database to predict nuclear reactor types using Factor Analysis (FA) and various machine learning classification algorithms. The results reveal that FA is an effective method for dimension reduction and visualization. The FA-KNN, Random Forest (RF), and Multilayer Perceptron (MLP) algorithms are applied using a consistent dataset partition to ensure unbiased comparisons. The prediction results based on 10-fold stratified cross-validation are quite promising and the Receiver Operating Characteristic (ROC) curves for multi-class classification confirm the excellent generalization ability of models. Therefore, the application of machine learning techniques is highly effective for reactor-type forensics analysis, especially for RF and MLP.

B-122 Improving the Diagnostic Utility of Multiplexed Assays in Laboratory Medicine with Machine Learning

Abstract Background The application of machine learning (ML) and artificial intelligence (AI) in precision medicine is of great interest and continues to grow. ML approaches offer a flexible and data-centric methodology for modeling diagnostic and prognostic trends, capable of identifying complex non-linear patterns and relationships between predictor variables and outcomes within large data sets. Although ML holds great promise, there are several common pitfalls that must be avoided for ML models to provide clinical utility. In the current study, we provide a rigorous and systematic evaluation of the utility of ML algorithms to predict disease in several representative data sets generated via multiplexed biomarker assays. We compare five major classes of ML algorithms compared to more traditional combinatorial analysis (logistic regression) to offer perspective against the current gold standard. Methods We obtained five data sets from published literature that include measurements of multiplexed biomarkers in patients with obstructive sleep apnea, sepsis, and breast and colon cancers in comparison to healthy controls. These datasets were generated by bead-based and planar multiplexed immunoassays. We chose representative ML algorithms from the classes of support vector machines (SVM), random forests (RF), neural networks (NNet), and extreme gradient boosting (xgBoost) in comparison to logistic regression (LR). Datasets were randomly split into internal validation and external validation sets via 10 iterations of permutation analysis. We tuned and validated the models using 5-fold cross validation, then tested those on external subsets. We report the average of several model classification metrics and associated variances across all resampled data sets. Results For the classification of sepsis using inflammatory cytokines measured on the Luminex bead-based immunoassay, RF outperformed LR in external validation sets by a 9.8% increase in AUROC. For classification of breast cancer patients, xgBoost and RF performed best with 20.0% and 16.9% increases in AUROC, respectively, whereas for classification of colon cancer patients, NNet outperformed LR by 11.5%. Across all datasets, on average, all ML algorithms outperformed LR in the internal validation sets by a 7.9% (6.3% - 9.9%) increase in AUROC. However, in external validation sets, only xgBoost and SVM outperformed LR, where RF and NNet showed signs of overfitting. Conclusions Overall, certain ML algorithms show improvement over LR for diagnostic applications using multiplexed assays, by up to 20% increase in AUROC. Interpretation of ML results must be performed rigorously to prevent overly optimistic and unrealistic conclusions. In the context of laboratory medicine, our results showcase the utility of ML in clinical applications, while highlighting potential disadvantages of ML. This analysis underscores the need for establishing rigorous guidelines to support data analytics during the development of novel multiplexed assays, which is needed to advance precision medicine.

B-117 Rapid Classification of Cleanliness Grades of Vaginal Discharge through Computational Analysis of SERS Spectra via Variational Auto-Encoder Model And Long Short-Term Memory Networks

Abstract Background Classification of cleanliness grades of vaginal discharge plays a pivotal role in clinical laboratories, serving as indicators for the presence of various vaginal infections such as bacterial vaginosis and candidal vulvovaginitis, etc. The conventional method for assessing vaginal discharge cleanliness involves manual microscopy, a technique susceptible to variations based on the examiner's expertise and subjective judgment, hindering the attainment of swift and accurate diagnoses. Methods This study introduces a novel approach by integrating Surface-Enhanced Raman Spectroscopy (SERS) with a deep learning algorithm for the rapid classification of vaginal discharge cleanliness grades. The proposed classification method combines the Variational Auto-Encoder (VAE) and Long Short-Term Memory (LSTM) to process SERS spectra. The quality of the training data is assessed using the signal-to-noise ratio (SNR) as a standard indicator. Results The application of the VAE-LSTM algorithm significantly enhances classification performance, providing swift and accurate results while reducing the required time. Comparative analyses with various classical machine learning algorithms and evaluation metrics reveal the superior predictive capability and time efficiency of the VAE-LSTM algorithm, demonstrating 100% accuracy and a fitting time of 2 seconds. Blind testing of SERS spectra collected from unknown patients further validates the reliability of the proposed method. Conclusions This study presents a promising and practical approach for the rapid classification of vaginal discharge cleanliness grades. The integration of the SERS technique with the VAE-LSTM algorithm not only ensures accuracy and efficiency in clinical diagnosis but also lays the foundation for advancements in non-invasive diagnostic technologies in gynecological healthcare.

An AI-Based Approach for Developing a Recommendation System for Underground Mining Methods Pre-Selection

Selecting the most appropriate mining method to recover mineral resources is a critical decision-making task in mining project development. This study introduces an artificial intelligence-based mining methods recommendation system (AI-MMRS) for the pre-selection of underground mining methods. The study integrates and evaluates the capability of two approaches for mining methods selection (MMS): the memory-based collaborative filtering (CF) approach aided by the UBC-MMS system to predict the top-3 relevant mining methods and supervised machine learning (ML) classification algorithms to enhance the effectiveness and novelty of the AI-MMRS, addressing the limitations of the CF approach. The results reveal that the memory-based CF approach achieves an accuracy ranging from 81.8% to 87.9%. Among the classification algorithms, artificial neural network (ANN) and k-nearest neighbors (KNN) classifiers perform the best, with accuracy levels of 66.7% and 63.6%, respectively. These findings demonstrate the effectiveness and viability of both approaches in MMS, acknowledging their limitations and the need for continuous training and optimization. The proposed AI-MMRS for the pre-selection stage supplemented by the direct involvement of mining professionals in later stages of MMS, has the potential to significantly aid in the MMS decision-making, providing data-driven and experience-based recommendations following the ongoing evolution of mining practices.

Weighted naïve bayes multi-user classification for adaptive authentication

Abstract Machine learning classification algorithms have been extensively utilized in addressing user authentication challenges. Nonetheless, a majority of solutions categorize users into three classes, whereas adaptive authentication scenarios necessitate classification beyond this threshold. The rationale behind this limitation has not been thoroughly explored. The current study leveraged the Naive Bayes theorem for user authentication endeavors to assess the risk associated with login attempts. The Naive Bayes Machine Learning algorithm, along with its variations such as Gaussian, Categorical, and Bernoulli, was applied on both weighted and unweighted datasets to ascertain risk levels and categorize them into six classes. Additionally, the classification task was executed using alternative algorithms. The outcomes of cross-validation and comparative analyses revealed that the performance was commendable for up to three classes, after which a decrease was observed in certain Naive Bayes and SVM classifiers. Among the Naïve Bayes family, the Bernoulli NB algorithm exhibited superior performance but was surpassed by Decision Trees, SVM, XGB, and Random Forests. Notably, the weighted dataset consistently outperformed the unweighted counterpart, with the allocation of weights significantly influencing algorithmic efficacy. The 80:20 split strategy yielded the most favorable outcomes in contrast to the 70:30 and 60:40 splits, albeit no significant variances were detected during cross-validation. Non-Naïve Bayes algorithms demonstrated superior performance compared to Naïve Bayes algorithms. For Naïve Bayes, optimal performance is achieved with three classes, highlighting its utility in conditional risk calculation, while non-Naïve Bayes multi-classification algorithms are more suitable for classification tasks due to the problem’s inherent compatibility with conditional probabilities. In conclusion, it is imperative to acknowledge that the characteristics of the data, the use of weights, and the data splitting methodology significantly influence the accuracy of machine learning algorithms in multi-class user classification.

A disrupted learning mechanism in standard quantum systems followed by their self-organizing process

Recently, the fusion between quantum mechanics and machine learning has gained much attention, where classical machine learning algorithms are adapted for quantum computers to significantly amplify data analysis by leveraging the unique effects of quantum reality. In this short paper, by focusing on the quantum trajectories of particles, we find that under general requirements, quantum systems follow a disrupted version of the gradient descent model, a basic machine learning algorithm, where the learning is distorted due to the self-organizing process of the quantum system. Such a learning process is possible only when we assume dissipation, i.e., that the quantum system is open. The friction parameter determines the nonlinearity of the quantum system. We then provide an empirical demonstration of the proposed model.

A Machine Learning Predictive Model for Ureteroscopy Lasertripsy Outcomes in a Pediatric Population-Results from a Large Endourology Tertiary Center.

Introduction: We aimed to develop machine learning (ML) algorithms for the automated prediction of postoperative ureteroscopy outcomes for pediatric kidney stones based on preoperative characteristics. Materials and Methods: Data from pediatric patients who underwent ureteroscopy for stone treatment by a single experienced surgeon, between 2010 and 2023 in Southampton General Hospital, were retrospectively collected. Fifteen ML classification algorithms were used to investigate correlations between preoperative characteristics and postoperative outcomes: primary stone-free status (SFS, defined as stone fragments <2 mm at the end of the procedure confirmed endoscopically and no evidence of stone fragments >2 mm at Xray kidney-ureters-bladder (XR KUB) or ultrasound kidney-ureters-bladder (US KUB) at 3 months follow-up) and complications. For the task of complication and stone status, an ensemble model was made out of Bagging classifier, Extra Trees classifier, and linear discriminant analysis. Also, a multitask neural network was constructed for the simultaneous prediction of all postoperative characteristics. Finally, explainable artificial intelligence techniques were used to explain the prediction made by the best models. Results: The ensemble model produced the highest accuracy (90%) in predicting SFS, finding correlation with overall stone size (-0.205), presence of multiple stones (-0.127), and preoperative stenting (-0.102). Complications were predicted by Synthetic Minority Oversampling Technique (SMOTE) oversampled dataset (93.3% accuracy) with relation to preoperative positive urine culture (-0.060) a1nd SFS (0.003). Training ML for the multitask model, accuracies of 83.3% and 80% were respectively reached. Conclusion: ML has a great potential of assisting health care research, with possibilities to investigate dataset at a higher level. With the aid of this intelligent tool, urologists can implement their practice and develop new strategies for outcome prediction and patient counseling and informed shared decision-making. Our model reached an excellent accuracy in predicting SFS and complications in the pediatric population, leading the way to the validation of patient-specific predictive tools.

Towards reliable hyperspectral imaging biomarkers of CT26 murine tumor model

The non-invasive monitoring of tumor growth can offer invaluable diagnostic insights and enhance our understanding of tumors and their microenvironment. Integrating hyperspectral imaging (HSI) with three-dimensional optical profilometry (3D OP) makes contactless and non-invasive tumor diagnosis possible by utilizing the inherent tissue contrast provided by visible (VIS) and near-infrared (NIR) light. Consequently, valuable information regarding tumors and healthy tissues can be extracted from the acquired hyperspectral images. Until now, very few methods have been used to monitor tumor models in vivo daily and non-invasively. In this research, we conducted a 14-day study monitoring BALB/c mice with subcutaneously grown CT26 murine colon carcinomas in vivo, commencing on the day of tumor cell injection. We extracted physiological properties such as total hemoglobin (THB) and tissue oxygenation (StO2) using the inverse adding-doubling (IAD) algorithm and manually segmented the tissues. We then selected the ten most relevant features describing tumors using the Max-Relevance Min-Redundancy (MRMR) algorithm and utilized 30 classic and advanced machine learning (ML) algorithms to discriminate tumors from healthy tissues. Finally, we tested the robustness of feature selection and model performance by smoothing tissue parameter maps extracted by IAD with a variable kernel and omitting selected training data. We could discriminate CT26 tumor models from surrounding healthy tissues with an area under the curve (AUC) of up to 1 for models based on the gradient boosting method, linear discriminant analysis, and random forests. Our findings help pave the way for precise and robust imaging biomarkers that could aid tumor diagnosis and advance clinical practice.

Comparative analysis of machine learning algorithms for sentiment classification in social media text

In the era of social media, sentiment analysis has become crucial for understanding public opinion. This study presents a comparative analysis of five machine learning algorithms for sentiment classification in social media text: Logistic Regression, Support Vector Machines (SVM), Random Forest, Naive Bayes, and Gradient Boosting. Using a dataset of 100,000 tweets collected over three months, we evaluated these algorithm's performance in classifying sentiments as positive, negative, or neutral. The data underwent extensive preprocessing, including cleaning, normalization, and addressing class imbalance using SMOTE. Our results show that Logistic Regression and SVM achieved the highest overall accuracy at 86.22%, demonstrating balanced performance across all sentiment classes. Random Forest followed closely with 82.59% accuracy, while Naive Bayes and Gradient Boosting showed lower but still noteworthy performance at 70.45% and 69.96% respectively. All models exhibited challenges in classifying negative sentiments, suggesting potential areas for improvement. The study provides insights into each algorithm's strengths and weaknesses, offering guidance for practitioners in selecting appropriate methods for sentiment analysis tasks. Our findings contribute to the ongoing research in applying machine learning to the complex task of sentiment analysis in the rapidly evolving landscape of social media communication.

Battery Condition Monitoring of Quadrotor UAV Using Machine Learning Classification Algorithm

Unmanned aerial vehicle flight performance and efficiency rely on various factors. Flight instabilities can happen due to malfunctions inside the system and disturbances from the external environment. Battery status plays a significant role in healthy flight conditions. A weak battery will affect the performance of propellers and motors, and the presence of wind disturbance can contribute towards inefficient flying capabilities. Therefore, investigation of fault at the early stage is crucial to maintain the great performance of the UAV. This paper aims to investigate the best prediction system from the existing machine learning algorithm such as Decision Tree (DT), Linear Discriminant (LD), Naïve Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN) and Neural Network (NN) to classify the battery condition of the quadrotor by extracting the features from the displacement time series dataset. By using recorded flight data, it will be statistically analyzed to extract the flying condition features. The extracted features are the Euclidian distance (ED), speed, acceleration, Periodogram Power spectral density (PSD) and Fast Fourier Transform (FFT) of the signal. The result shows that the two best classifier algorithms are the Decision Tree and Neural Network models with training accuracy of 98% and 93% in Set A and B, respectively.

Comparing Machine Learning Regression Models for Early-Age Compressive Strength Prediction of Recycled Aggregate Concrete

A branch of artificial intelligence called machine learning is well-positioned as a prediction method that can take into consideration several influencing factors and complex inter-factor connections. Without being specifically trained to do so, these machine learning models have the ability to generalize, predict, and learn from data. Regression theory is a key topic in statistical modelling and machine learning. The main goal of this study is to compare the performance of several popular machine learning regression models for predicting the early-age compressive strength of concretes made from recycled concrete aggregates from a structure that demolished following the Sivrice-Elazig earthquake on January 24, 2020. Early-age concrete compressive strength is even more crucial due to factors like the fact that there are thousands of newly built structures in the aftermath of the earthquake, the quick manufacturing of these structures, and the completion of the project in the lowest amount of time. Determining the early-age concrete strength with high accuracy and in a useful manner is crucial for this reason. Seven different classical machine learning algorithms were employed in this study to achieve all of these goals. Early-age concrete compressive strength values were considered for 1 and 3 days. The relationship between the experimental results and the predicted outcomes of the employed algorithms was investigated, and a thorough comparison of these intelligent regression algorithms was conducted. Within the scope of sustainable development and circular economy goals, it is thought that this article will make significant contributions to the literature in terms of utilizing these waste materials and determining the early-age compressive strengths of the concretes produced with high accuracy.

Retracted: A Pre-Hospital Prediction Tool for Stroke Type Utilising Machine Learning Techniques

Abstract Background Stroke is a leading cause of death worldwide and leaves most survivors with permanent disability. 12.2 million people suffer strokes annually, of which 70% occur in low- and middle-income countries (LMIC). Stroke presents diagnostic challenges which limits access to effective, but time-sensitive, reperfusion therapies. Currently, stroke is diagnosed by the combination of clinical symptoms and signs and brain imaging. The aim of this study was to develop clinical prediction tools, for differentiating acute ischemic stroke (AIS) and intracerebral haemorrhage (ICH) in the pre-hospital setting without neuroimaging, utilising machine learning (ML). Methods We included 5349 patients with confirmed acute stroke who presented within 6 hours of symptom onset, from the INTERSTROKE case-control study. We compared predictions using six commonly used ML algorithms for classification i.e. multivariate logistic regression (MLR), k-nearest neighbour (KNN), support vector machine (SVM), neural networks (NN), random forests (RF), XGBoost. Algorithm inputs included medical history, symptoms and signs, and point-of-care blood tests, with an output of final diagnosis. Results The training cohort included 4280 patients with 3084 ischemic stroke and 1196 ICH cases, and the test cohort included 1069 patients with 771 ischemic stroke and 298 ICH cases. From 174 potential variables, we used univariate analyses to select 52 candidate variables for inclusion in the ML algorithms. These were selected based on their ability to predict (p &amp;lt; 0.20) stroke subtype. The AUCs in the test cohort for each of the ML classification models were: MLR 0.819, KNN 0.795, SVM 0.802, NN 0.814, RF 0.814, and XGBoost 0.807. Conclusion ML models showed great potential to predict stroke type at the prehospital stage and have comparable predictive abilities to previously developed ML prediction tools. This may facilitate point-of-care decisions and management of stroke in the pre-hospital setting in the future.

Machine learning for anxiety and depression profiling and risk assessment in the aftermath of an emergency

Background & objectivesMental health disorders pose an increasing public health challenge worsened by the COVID-19 pandemic. The pandemic highlighted gaps in preparedness, emphasizing the need for early identification of at-risk groups and targeted interventions. This study aims to develop a risk assessment tool for anxiety, depression, and self-perceived stress using machine learning (ML) and explainable AI to identify key risk factors and stratify the population into meaningful risk profiles. MethodsWe utilized a cohort of 9291 individuals from Northern Spain, with extensive post-COVID-19 mental health surveys. ML classification algorithms predicted depression, anxiety, and self-reported stress in three classes: healthy, mild, and severe outcomes. A novel combination of SHAP (SHapley Additive exPlanations) and UMAP (Uniform Manifold Approximation and Projection) was employed to interpret model predictions and facilitate the identification of high-risk phenotypic clusters. ResultsThe mean macro-averaged one-vs-one AUROC was 0.77 (± 0.01) for depression, 0.72 (± 0.01) for anxiety, and 0.73 (± 0.02) for self-perceived stress. Key risk factors included poor self-reported health, chronic mental health conditions, and poor social support. High-risk profiles, such as women with reduced sleep hours, were identified for self-perceived stress. Binary classification of healthy vs. at-risk classes yielded F1-Scores over 0.70. ConclusionsCombining SHAP with UMAP for risk profile stratification offers valuable insights for developing effective interventions and shaping public health policies. This data-driven approach to mental health preparedness, when validated in real-world scenarios, can significantly address the mental health impact of public health crises like COVID-19.