Classification-based Machine Learning Research Articles

Beyond network connectivity: A classification approach to brain age prediction with resting-state fMRI

The brain is a complex, dynamic organ that shows differences in the same subject at various periods. Understanding how brain activity changes across age as a function of the brain networks has been greatly abetted by fMRI. Canonical analysis consists of determining how alterations in connectivity patterns (CPs) of certain regions are affected. An alternative approach is taken here by not considering connectivity but rather features computed from recordings at the regions of interest (ROIs). Using machine learning (ML) we assess how neural signals are altered by and prospectively predictive of age and sex via a methodology that is novel in drawing upon pairwise classification across six decades of subjects’ chronological ages. ML is used to answer the equally important questions of what properties of the computed features are most predictive as well as which brain networks are most affected by aging. It was found that there is decreased differentiation among the neural signals of older subjects that are separated in age by the same number of years as younger subjects. Furthermore, the burstiness of the signals change at different rates between males and females. The findings provide insight into brain aging via an ROI-based analysis, the consideration of several feature groups, and a novel classification-based ML pipeline. There is also a contribution to understanding the effects of data aggregated from different recording centers on the conclusions of fMRI studies.

Certain Investigations on Feature Selection Technique Using Artificial Immune Systems for EEG Color Visualization Classification

This paper aims to extract and select the significant features of electroencephalogram (EEG) signals and classify the visual stimulation of distinct colors. In this work, a novel method for selecting distinct colors using EEG signals called affinity artificial immune and Daubechies wavelet time-based learning (AAIDWTL) is proposed. Initially, the EEG signals were collected in a controlled environment and an in-built band-pass filter was applied to remove the artifacts. The filtered signals were converted into frequency domain signals using least squarebased short-term Fourier transform. After that, by utilizing Daubechies wavelet statistical time-based feature extraction model the time domain features were extracted. Followed by, computationally efficient features were selected using an affinity artificial immune-based feature selection model. The selected features were classified using a polynomial kernel multiclass classification-based machine learning algorithm and achieved an accuracy of 97.5% when compared with other methods like linear discriminant analysis (LDA) which obtained only 92%. Furthermore, while utilizing the proposed method classification time was considerably less when compared to LDA. The experimental result shows that the proposed color stimulation of the EEG signals method achieved greater improvement in terms of both classification time and classification accuracy with a minimum false positive rate.

Machine learning vs. traditional regression analysis for fluid overload prediction in the ICU

Fluid overload, while common in the ICU and associated with serious sequelae, is hard to predict and may be influenced by ICU medication use. Machine learning (ML) approaches may offer advantages over traditional regression techniques to predict it. We compared the ability of traditional regression techniques and different ML-based modeling approaches to identify clinically meaningful fluid overload predictors. This was a retrospective, observational cohort study of adult patients admitted to an ICU ≥ 72 h between 10/1/2015 and 10/31/2020 with available fluid balance data. Models to predict fluid overload (a positive fluid balance ≥ 10% of the admission body weight) in the 48–72 h after ICU admission were created. Potential patient and medication fluid overload predictor variables (n = 28) were collected at either baseline or 24 h after ICU admission. The optimal traditional logistic regression model was created using backward selection. Supervised, classification-based ML models were trained and optimized, including a meta-modeling approach. Area under the receiver operating characteristic (AUROC), positive predictive value (PPV), and negative predictive value (NPV) were compared between the traditional and ML fluid prediction models. A total of 49 of the 391 (12.5%) patients developed fluid overload. Among the ML models, the XGBoost model had the highest performance (AUROC 0.78, PPV 0.27, NPV 0.94) for fluid overload prediction. The XGBoost model performed similarly to the final traditional logistic regression model (AUROC 0.70; PPV 0.20, NPV 0.94). Feature importance analysis revealed severity of illness scores and medication-related data were the most important predictors of fluid overload. In the context of our study, ML and traditional models appear to perform similarly to predict fluid overload in the ICU. Baseline severity of illness and ICU medication regimen complexity are important predictors of fluid overload.

Controlling factor for fracture resistance and ionic conduction in glassy lithium borophosphate electrolytes

Glasses are promising candidates as solid electrolytes for all-solid-state batteries due to their isotropic ionic conduction, formability, as well as high chemical, thermal and electrochemical stability. However, their mechanical properties and ionic conductivity need to be improved. Here, based on molecular dynamics simulations and classification-based machine learning, we reveal that both fracture behavior and ionic conduction in glassy lithium borophosphate electrolytes are encoded in their static structures. By systematically varying the Li and B content, we demonstrate that the machine learning-based structural descriptor termed “softness” can be used as an indicator for both fracture resistance and ionic conductivity. The “softness” metric is calculated from the static local atomic environment, but well captures the long-term dynamics of individual atoms. Notably, the propensities for B atoms to undergo bond-switching (correlated with fracture) and for Li ions to migrate (correlated with ionic conductivity) increase with an increase in atomic softness. Specifically, the out-of-equilibrium interaction of B and Li with oxygen neighbors enhances the propensity for B and Li to undergo bond-switching or rearrangement when experiencing stimuli. These results enable finding the optimum chemical compositions for glassy solid electrolytes with high mechanical stability and high ionic conductivity.

Machine Learning-Based Prediction of Distant Recurrence in Invasive Breast Carcinoma Using Clinicopathological Data: A Cross-Institutional Study.

Breast cancer is the most common type of cancer worldwide. Alarmingly, approximately 30% of breast cancer cases result in disease recurrence at distant organs after treatment. Distant recurrence is more common in some subtypes such as invasive breast carcinoma (IBC). While clinicians have utilized several clinicopathological measurements to predict distant recurrences in IBC, no studies have predicted distant recurrences by combining clinicopathological evaluations of IBC tumors pre- and post-therapy with machine learning (ML) models. The goal of our study was to determine whether classification-based ML techniques could predict distant recurrences in IBC patients using key clinicopathological measurements, including pathological staging of the tumor and surrounding lymph nodes assessed both pre- and post-neoadjuvant therapy, response to therapy via standard-of-care imaging, and binary status of adjuvant therapy administered to patients. We trained and tested four clinicopathological ML models using a dataset (144 and 17 patients for training and testing, respectively) from Duke University and validated the best-performing model using an external dataset (8 patients) from Dartmouth Hitchcock Medical Center. The random forest model performed better than the C-support vector classifier, multilayer perceptron, and logistic regression models, yielding AUC values of 1.0 in the testing set and 0.75 in the validation set (p < 0.002) across both institutions, thereby demonstrating the cross-institutional portability and validity of ML models in the field of clinical research in cancer. The top-ranking clinicopathological measurement impacting the prediction of distant recurrences in IBC were identified to be tumor response to neoadjuvant therapy as evaluated via SOC imaging and pathology, which included tumor as well as node staging.

Predicting HIV infection in the decade (2005-2015) pre-COVID-19 in Zimbabwe: A supervised classification-based machine learning approach.

The burden of HIV and related diseases have been areas of great concern pre and post the emergence of COVID-19 in Zimbabwe. Machine learning models have been used to predict the risk of diseases, including HIV accurately. Therefore, this paper aimed to determine common risk factors of HIV positivity in Zimbabwe between the decade 2005 to 2015. The data were from three two staged population five-yearly surveys conducted between 2005 and 2015. The outcome variable was HIV status. The prediction model was fit by adopting 80% of the data for learning/training and 20% for testing/prediction. Resampling was done using the stratified 5-fold cross-validation procedure repeatedly. Feature selection was done using Lasso regression, and the best combination of selected features was determined using Sequential Forward Floating Selection. We compared six algorithms in both sexes based on the F1 score, which is the harmonic mean of precision and recall. The overall HIV prevalence for the combined dataset was 22.5% and 15.3% for females and males, respectively. The best-performing algorithm to identify individuals with a higher likelihood of HIV infection was XGBoost, with a high F1 score of 91.4% for males and 90.1% for females based on the combined surveys. The results from the prediction model identified six common features associated with HIV, with total number of lifetime sexual partners and cohabitation duration being the most influential variables for females and males, respectively. In addition to other risk reduction techniques, machine learning may aid in identifying those who might require Pre-exposure prophylaxis, particularly women who experience intimate partner violence. Furthermore, compared to traditional statistical approaches, machine learning uncovered patterns in predicting HIV infection with comparatively reduced uncertainty and, therefore, crucial for effective decision-making.

Prediction of Survival of Heart Failure Patients: An Application of Classification-Based Machine Learning Algorithms

Kardiyo-vasküler hastalıklar dünya genelinde en çok ölüme sebep olan hastalıklar arasında yer &#x0D; almaktadır. Bir kardiyo-vasküler hastalık olan kalp yetmezliği, kalbin vücudun ihtiyaç duyduğu kanı &#x0D; pompalayamaması durumudur. Ülkemizde sıklıkla görülen bu hastalığın sonucu olarak ölümler &#x0D; yaşanmaktadır. Bu çalışmada kalp yetmezliğe sahip hastaların sağ kalım veya ölüm durumlarının tahmin &#x0D; edilmesi için makine öğrenmesi tabanlı bir yaklaşım önerilmektedir. Üç farklı sınıflandırma algoritması &#x0D; kullanılarak önerilen yöntemin etkinliği değerlendirilmektedir. Gerçekleştirilen deneylerde, Yapay Sinir &#x0D; Ağı algoritması ile en yüksek doğruluk değerine (86.67%) ulaşılmıştır. Önerilen yöntem, ölüm riskinin &#x0D; yüksek olduğu kalp yetmezliği hastalarına daha etkin ve uygun tedavi planlarının hazırlanması açısından &#x0D; yol gösterici olacaktır.

Abstract LB070: Predicting distant recurrences in invasive breast carcinoma patients using their clinicopathological profiles

Abstract Purpose: To determine whether the classification-based machine learning (ML) or artificially intelligent (AI) techniques can predict distant recurrence flags (yes or no) in invasive breast cancer patients using the data comprising several clinicopathological measurements such as pathological staging of tumor and surrounding nodes deemed both pre and post neoadjuvant (i.e. either chemo, radiation or hormone based) therapy including imaging based therapy-response as well and the status of adjuvant therapy (Chemo or Anti-Her2/Neu antibody) therapy were administered to patients post resection of their tumor to minimize the possibility of diseases-recurrence in them. Method: The clinicopathological data from a retrospective study of 900 breast cancer patients at Duke university (posted at The Cancer Imaging Archive, i.e. TCIA) was tapped for our study. These patients received neoadjuvant therapies such as Chemo-, Radiation or Endocrine Hormone based therapy, had their responses evaluated through imaging and pathological staging (both tumor and nodal), and received Adjuvant therapies (such as Chemo or Anti-Her2/Neu antibody therapy) post tumor resection. The patient entries pertaining to ungraded tumor responses from therapies or those labelled as ‘Not applicable (NA)’ for any of the aforementioned clinicopathological parameters were filtered out retaining 161 patient-entries which were further split into train and test sets in 90:10 ratio where 90% were used for training (ntrain=144) and 10% for testing (ntest=17) of several ML models. Classification based machine learning models such as Random Forest (RF), C-Support Vector Classification (SVC) and Supervised Neural Network (aka Multi-Layer Perceptron, i.e. MLP) were employed to train and test aforementioned clinicopathological parameters to predict distant recurrence flags or labels (i.e. Yes or No). The training of the model was conducted using ImaGene software. Out of the three aforementioned models, RF seemed to have performed best with grid search activated over tree-depth hyperparameter to predict distant recurrence flags in test patients (ntest=17) at AUC=1.0 (p&amp;lt;0.002). Further, the validation of the model was conducted using external clinicopathological dataset of 20 patients from Dartmouth-Hitchcock (DH) Medical center which yielded AUC&amp;gt;0.75. Conclusion: Random Forest model trained using grid search through tree-depth hyperparameter can predict distant recurrence flags (Yes or No) in invasive breast cancer patients at AUC in the range of 0.75 1.0 across multiple institutions, thereby aiding portability of models and advancing the idea of multi-instutional ML/AI model validations in Invasive Breast Carcinoma subtype of Breast Cancer. Citation Format: Shrey S. Sukhadia, Kristen Muller, Adrienne Workman, Shivashankar Nagaraj, Olivier Gevaert. Predicting distant recurrences in invasive breast carcinoma patients using their clinicopathological profiles [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 2 (Clinical Trials and Late-Breaking Research); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(8_Suppl):Abstract nr LB070.

Machine Learning Approach for Prediction of the Online User Intention for a Product Purchase

The deployment of self-learning computer algorithms that can automatically enhance their performance via experience is referred to as machine learning in ecommerce and is a crucial trend of the retail digital transformation. Machine learning algorithms can be unambiguously trained by analysing big datasets, identifying repeating patterns, relationships, and anomalies among all of this data, and creating mathematical models resembling such associations. These models are improved when the algorithms analyse ever-increasing amounts of data, providing us with useful insights into specific ecommerce-related events and the links between all the variables that underlie them. A tool that has been quite effective in studying current affairs, predicting future trends, and making data-driven decisions. The present work investigates the implementation of machine learning algorithms to predict the user intention for purchasing a product on a specific store's website. An Online Shoppers Purchasing Intention data set from the UC Irvine Machine Learning Repository was used for this investigation. In this study, two classification-based machine learning algorithms i.e. Stochastic Gradient Descent (SGD) algorithm and Random Forest algorithm were used. SGD algorithm was used for first time in prediction of the online user intention. The results showed that the Random Forest resulted in the highest F1-Score of 0.90 in contrast to the Stochastic Gradient Descent algorithm.

Diagnosing Obesity using Classification based Machine Learning Models

Diagnosing Obesity using Classification based Machine Learning Models

Diagnosing obesity using classification-based machine learning models

Diagnosing obesity using classification-based machine learning models

One-class Classification-Based Machine Learning Model for Estimating the Probability of Wildfire Risk

Wildfires have caused devastating consequences to property and human and animal lives, which has become a global problem. Consequently, advanced wildfire prediction models are required to treat complex features and climate conditions. As a result, Machine Learning and Deep Learning models are becoming popular. However, creating a balanced true and false labeled dataset in the wild-fire domain is often challenging. Hence, One-class classification models are a promising approach to overcome this concern. In this paper, several One-class classification models are investigated; linear models (Principal Component Analysis and One-Class Support Vector Machines), outlier ensemble models (Lightweight On-line Detector of Anomalies and Locally Selective Combination of Parallel Outlier Ensembles), proximity-based models (Histogram-based Outlier Score and Rotation-based Outlier Detection), probabilistic models (Unsupervised Outlier Detection Using Empirical Cumulative Distribution Functions and Copula-Based Outlier Detection), neural network-based models (Deep One-class Classification and Adversarially Learned Anomaly Detection) is used in two case studies for California and Western Australian states. In conclusion, it was found that Deep learning-based One-class classification models outperform other models in terms of performance and feature importance of showcasing the effectiveness of deep neural network models in the wildfire prediction domain.

The Design and Implementation Of A Machine Learning Based Breast Cancer Diagnostic System

The adoption of information technology in the Ghanian health sector remains for the most part limited to the Greater Accra region. The technologies adopted are mainly used for communication, telemedicine, record keeping and database systems. Advanced analytical technologies like machine learning and artificial intelligence can greatly improve the Ghanian health. But despite their great potential, they are still unpopular. The current project is on the use of classification-based machine learning technology for breast cancer disease diagnosis. This project is designed and implemented as an online based platform to be used mainly by laboratories agents and researchers. This system was built to simplify the diagnostic process for the disease and increase its accuracy level. Keywords: Machine Learning, K-Nearest Neighbor and eHealth, Breast Cancer, Diagnostics

Deciphering a structural signature of glass dynamics by machine learning

The dynamics of atoms plays a key role in governing various dynamical and transport properties of glasses. However, it remains elusive which structural features (if any) control atom dynamics in glasses. Here, based on million-atom molecular dynamics simulations and classification-based machine learning, we extract a needle in a haystack by identifying a local, nonintuitive structural signature (a revised version of the recently developed softness metric) that governs glass dynamics. We do so by investigating the ion mobility in sodium silicate glasses---a realistic, archetypal glass---finding that the sodium ion mobility is largely encoded in its initial softness, wherein softer Na atoms exhibit higher mobility. Importantly, our approach allows us to interpret the machine-learned softness metric and thus elucidate the atomistic origin of the ion mobility. Namely, we find that Na mobility is anticorrelated with the local density of defect oxygen neighbors that are located between the nearest two coordination shells. This local packing order offers a potential path to develop glass formulations with tailored dynamical properties. Finally, we demonstrate that the softness is strongly anticorrelated with the activation energy for Na atom reorganization.

Machine learning application to layer counting in speleothems

Some speleothems, in particular stalagmites, are laminated at the visible and microscopic scale, with the latter visible using fluorescence microscopy (e.g., confocal laser scanning microscopy). These laminations can be used to supplement speleothem chronologies, although this process is laborious and lateral variations in lamination geometry and quality necessitate a detailed look over an entire scan as opposed to a simple one-dimensional transect. In order to assist this process, we develop a classification-based machine learning algorithm using an open-source machine learning package. This algorithm is optimized for stalagmites growing at 20–100 μm yr−1 and outputs a 2-dimensional layer density map which may aid in quantitatively interpreting past variations in speleothem growth rate. This algorithm requires user supervision and interpretation, as image artefacts and magnification settings may complicate model output.

A classification and regression tree algorithm for heart disease modeling and prediction

Heart disease remains the leading cause of death, such that nearly one-third of all deaths worldwide are estimated to be caused by heart-related conditions. Advancing applications of classification-based machine learning to medicine facilitates earlier detection. In this study, the Classification and Regression Tree (CART) algorithm, a supervised machine learning method, has been employed to predict heart disease and extract decision rules in clarifying relationships between input and output variables. In addition, the study’s findings rank the features influencing heart disease based on importance. When considering all performance parameters, the 87% accuracy of the prediction validates the model’s reliability. On the other hand, extracted decision rules reported in the study can simplify the use of clinical purposes without needing additional knowledge. Overall, the proposed algorithm can support not only healthcare professionals but patients who are subjected to cost and time constraints in the diagnosis and treatment processes of heart disease.

Enhancing the security in cyber-world by detecting the botnets using ensemble classification based machine learning

With various malware, botnets are the legitimate risk increasing against cybersecurity providing criminal operations like malware dispersal, distributed denial of service attacks, fraud clicking, phishing, and identification of theft. Existing techniques used for detection of botnet, which are suitable only for specific command of botnet and protocol for controlling and do not support botnet detection at earlier stages. In several computer security defense systems, honeypots are deployed successfully by security defenders. As honeypots can attract botnet compromises and expose spies in botnet membership and behaviors of the attacker, they are broadly employed in botnet defense. Thus, attackers whose role is to construct and maintain botnets have to determine honeypot trap avoiding methods. To handle the issues related to botnet attacks, machine learning techniques are used to support detection and prevent bot attacks. An Ensemble Classifier Algorithm with Stacking Process (ECASP) is proposed in this paper to select optimal features fed as input to the machine learning classifiers to estimate the botnet detection performance. As a result, the method achieves proposed achieves 94.08% accuracy, 86.5% sensitivity, 85.68% specificity, and 78.24% F-measure.

Assessing the factors affecting the perceived crossing speed of pedestrians and investigating the direct and indirect effects of crash risk perception on perceived crossing speed

Walking is the primary means of transportation. For assessing individual's health, travel behaviour and benchmarking service levels of pedestrian infrastructure, walking/crossing speed acts as a relevant indicator. In spite of this, there is a paucity of literature that has focused on the evaluation of the individual's perceived crossing speed and the factors that influence it. Based on the responses collected from 616 pedestrians, the present study adopts an exploratory approach to identify the factors that influence an individual's perception of crossing speed. Five different types of classification-based machine learning methods (decision trees, boosting, K-nearest neighbour, random forest, and support vector machine) are employed to study the influence of selected factors of an individual's perceived crossing speed at mid-block crossings. Furthermore, using mediation analysis, it was analyzed if the crash risk perception (CRP) of an individual had a significant effect on their perceived crossing speed (PCS) directly and indirectly through their discrete crossing pattern (CP). Study findings indicate that crossing pattern (frequency of crossing/walking for recreational trips per week), crash risk perception, and age were the three most factors influencing the respondent's perceived crossing speed (PCS). In comparison to other four classification-based machine learning methods, random forest proved to be the most accurate method. Mediation analysis indicated that crash risk perception (CRP) directly affects the perceived crossing speed of individuals (individuals tend to cross slowly if they perceive a high risk of crashes) as well as indirectly influences crossing speed through their crossing pattern (CP). This study contributes to the growing body of research in investigating relationship between crossing pattern, perceived safety, and walking speed of pedestrians. The results of the present study will be the basis for advocating the need to pay more attention to pedestrian safety perception, since perceived crossing speed and crossing pattern are directly related to it.

Combining machine learning with meta-analysis for predicting cytotoxicity of micro- and nanoplastics

A growing number of articles have been reported to investigate health consequences of micro- and nanoplastics (MNPs) exposure, but no coherent images have been achieved due to the incomparability between studies, the complexity and heterogeneity of the existing toxicity data of MNPs. Herein, this work developed a predictive modelling framework for the cytotoxicity of MNPs using classification-based machine learning approaches. A literature search led to 1824 sample points represented by 9 features describing physicochemical properties of MNPs, the cell-related attribute and experimental factors. The decision tree ensemble classifier built from all features (DTE1) showed strong predictive ability with an accuracy of 0.95, recall and precision of both 0.86. Feature selection was subsequently performed to identify the key ingredients that dominated the toxic properties of MNPs. A simplified classifier developed from 6 influencing features demonstrated a comparable model performance to DTE1. This result can help direct future studies toward better experimental design and report, facilitating the understanding of the pressing MNP-related health issues. With continuous integration of more representative research data, the developed model can be widely applicable to a spectrum of MNP cytotoxicity settings.

Deep learning models for predicting the survival of patients with chondrosarcoma based on a surveillance, epidemiology, and end results analysis.

BackgroundAccurate prediction of prognosis is critical for therapeutic decisions in chondrosarcoma patients. Several prognostic models have been created utilizing multivariate Cox regression or binary classification-based machine learning approaches to predict the 3- and 5-year survival of patients with chondrosarcoma, but few studies have investigated the results of combining deep learning with time-to-event prediction. Compared with simplifying the prediction as a binary classification problem, modeling the probability of an event as a function of time by combining it with deep learning can provide better accuracy and flexibility.Materials and methodsPatients with the diagnosis of chondrosarcoma between 2000 and 2018 were extracted from the Surveillance, Epidemiology, and End Results (SEER) registry. Three algorithms—two based on neural networks (DeepSurv, neural multi-task logistic regression [NMTLR]) and one on ensemble learning (random survival forest [RSF])—were selected for training. Meanwhile, a multivariate Cox proportional hazards (CoxPH) model was also constructed for comparison. The dataset was randomly divided into training and testing datasets at a ratio of 7:3. Hyperparameter tuning was conducted through a 1000-repeated random search with 5-fold cross-validation on the training dataset. The model performance was assessed using the concordance index (C-index), Brier score, and Integrated Brier Score (IBS). The accuracy of predicting 1-, 3-, 5- and 10-year survival was evaluated using receiver operating characteristic curves (ROC), calibration curves, and the area under the ROC curves (AUC).ResultsA total of 3145 patients were finally enrolled in our study. The mean age at diagnosis was 52 ± 18 years, 1662 of the 3145 patients were male (53%), and mean survival time was 83 ± 67 months. Two deep learning models outperformed the RSF and classical CoxPH models, with the C-index on test datasets achieving values of 0.832 (DeepSurv) and 0.821 (NMTLR). The DeepSurv model produced better accuracy and calibrated survival estimates in predicting 1-, 3- 5- and 10-year survival (AUC:0.895-0.937). We deployed the DeepSurv model as a web application for use in clinical practice; it can be accessed through https://share.streamlit.io/whuh-ml/chondrosarcoma/Predict/app.py.ConclusionsTime-to-event prediction models based on deep learning algorithms are successful in predicting chondrosarcoma prognosis, with DeepSurv producing the best discriminative performance and calibration.