Statistical Learning Techniques Research Articles

Impact of Diabetes Mellitus on Heart Failure Patients: Insights from a Comprehensive Analysis and Machine Learning Model Using the Jordanian Heart Failure Registry.

Heart failure (HF) is a common final pathway of various insults to the heart, primarily from risk factors including diabetes mellitus (DM) type 2. This study analyzed the clinical characteristics of HF in a Jordanian population with a particular emphasis on the relationship between DM and HF. This prospective study used the Jordanian Heart Failure Registry (JoHFR) data. Patients with HF were characterized by DM status and HF type: HF with preserved ejection fraction (HFpEF) or HF with reduced ejection fraction (HFrEF). Demographics, clinical presentations, and treatment outcomes were collected. Statistical analyses and machine learning techniques were carried out for the prediction of mortality among HF patients: Recursive Feature Elimination with Cross-Validation (RFECV) and Synthetic Minority Over-sampling Technique with Edited Nearest Neighbors (SMOTEENN) were employed. A total of 2007 patients with HF were included. Notable differences between diabetic and non-diabetic patients are apparent. Diabetic patients were predominantly male, older, and obese (p < 0.001 for all). A higher incidence of HFpEF was observed in the diabetes cohort (p = 0.006). Also, diabetic patients had significantly higher levels of cholesterol (p = 0.008) and LDL (p = 0.003), reduced hemoglobin levels (p < 0.001), and more severe renal impairment (eGFR; p = 0.006). Machine learning models, particularly the Random Forest Classifier, highlighted its superiority in mortality prediction, with an accuracy of 90.02% and AUC of 80.51%. Predictors of mortality included creatinine levels >115 µmol/L, length of hospital stay, and need for mechanical ventilation. This study underscores notable differences in clinical characteristics and outcomes between diabetic and non-diabetic heart failure patients in Jordan. Diabetic patients had higher prevalence of HFpEF and poorer health indicators such as elevated cholesterol, LDL, and impaired kidney function. High creatinine levels, longer hospital stays, and the need for mechanical ventilation were key predictors of mortality.

Exploring the Performance of Linear and Nonlinear Models of Time-of-Flight Secondary Ion Mass Spectrometry Spectra.

Multivariate statistical tools and machine learning (ML) techniques can deconvolute hyperspectral data and control the disparity between the number of samples and features in materials science. Nevertheless, the importance of generating sufficient high-quality sample replicates in training data cannot be overlooked, as it fundamentally affects the performance of ML models. Here, we present a quantitative analysis of time-of-flight secondary ion mass spectrometry (ToF-SIMS) spectra of a simple microarray system of two food dyes using partial least-squares (PLS, linear) and random forest (RF, nonlinear) algorithms. This microarray was generated by a high-throughput sample preparation and analysis workflow for fast and efficient acquisition of quality and reproducible spectra via ToF-SIMS. We drew insights from the bias-variance trade-off, investigated the performances of PLS and RF regression models as a function of training data size, and inferred the amount of data needed to construct accurate and reliable regression models. In addition, we found that the spectral concatenation of positive and negative ToF-SIMS spectra improved the model performances. This study provides an empirical basis for future design of high-throughput microarrays and multicomponent systems, for the purpose of analysis with ToF-SIMS and ML.

Cross-Language Translation and Comparative Study of English Literature Using Machine Translation Algorithms

Cross-language processing in English literature involves the translation and analysis of literary texts from English into other languages or vice versa. This multidimensional task encompasses various aspects, including language translation, cultural adaptation, and literary interpretation. Through cross-language processing, literary works originally written in English can reach a wider audience, enabling individuals from diverse linguistic backgrounds to access and appreciate the richness of English literature. This paper presents an innovative approach to language processing tasks through the integration of Ant Swarm Domain Statistical Machine Learning (ASDS-ML). Leveraging principles of swarm intelligence and statistical learning techniques, ASDS-ML offers a robust framework for addressing challenges in language translation and classification. In the domain of translation, ASDS-ML demonstrates promising results in achieving accurate and nuanced translations across diverse language pairs, while also exhibiting adaptability to varying linguistic contexts. Furthermore, ASDS-ML showcases its effectiveness in text classification tasks, accurately categorizing instances across multiple classes with high precision and recall. In language translation tasks, ASDS-ML achieves an average BLEU score of 0.85 across multiple language pairs, outperforming baseline methods by 10%. Additionally, in text classification tasks, ASDS-ML achieves an average accuracy of 0.92 across ten different classes, surpassing existing approaches by 5%.

Data to Decisions: Optimizing E-commerce Sales Potential with Analytics

Descriptive, diagnostic, predictive, and prescriptive analytics play crucial roles in optimizing the performance and user experience of an e-commerce site. Descriptive analytics involves examining historical data to gain insights into past performance, enabling businesses to identify trends, patterns, and anomalies. This analysis helps in understanding what has happened, such as identifying popular products or peak sales periods, and provides a foundation for further analysis. Diagnostic analytics goes beyond descriptive analytics by examining why certain events occurred, identifying factors that influenced outcomes, and uncovering strengths, weaknesses, and areas for improvement within an e-commerce platform. Predictive analytics utilizes statistical algorithms and machine learning techniques to forecast future trends and outcomes, enabling businesses to anticipate customer preferences, demand for specific products, and potential sales opportunities. By leveraging predictive insights, e-commerce sites can adjust strategies, inventory levels, and marketing campaigns proactively to stay ahead of the competition and meet evolving customer needs. Prescriptive analytics takes predictive insights to the next level by recommending specific actions or strategies to optimize business processes and achieve desired outcomes. This could involve personalized product recommendations, targeted marketing strategies, or dynamic pricing adjustments based on real-time data analysis. By harnessing these four types of analytics, e-commerce businesses can make informed, data-driven decisions, enhance customer experiences, drive sales growth, and maximize profitability.

Relationship Between Biophysical Properties of Antimicrobial Peptides (AMPs) and their Associated Drug Efficacies

Antibiotic resistance is a growing global health threat. One consequence is that patients with cystic fibrosis (CF) are prone to developing antibiotic resistant lung infections caused by multiple strains of bacteria, including Pseudomonas aeruginosa. Due to the limited number of treatment options for patients with chronic antibiotic resistant infections, there is a need for finding new antibiotics that allow for effective eradication of bacterial infections, such as those in the CF lung. Many antimicrobial peptides (AMPs) have been annotated in databases and are considered as potential alternatives for current antibiotics. However, in many instances, the suitability of AMPs as drug molecules has not been extensively explored. Here, we propose that certain molecular properties of AMPs favor high antibiotic efficacy. Using information from AMP databases, we combined statistical analyses and machine learning techniques to identify relationships between various biophysical properties of AMPs and their drug efficacies. Analyses from classification and regression trees (CART) and random forests suggest that net charge and maximum average hydrophobic moment are the most important properties in determining if a peptide is useful against P. aeruginosa infections in CF patients. Maximum average hydrophobic residue, average alpha helix propensity score, hydrophobic proportion, and peptide length still contribute to this determination but to lesser degrees. Cation-pi interactions, on the other hand, do not appear to factor into this decision at all. Based on these properties, our current work is focused on designing and experimentally testing new peptides that may have activity against P. aeruginosa infections.

A Novel &amp; Efficient LR LSTM AIS Route Data Prediction for Longer Range

The growth of technology has enabled different industries to generate an excessive amount of data- one such industry being the maritime sector. Sophisticated sensory systems installed on various vessels generate information at a very large scale which can further be used in optimizing operational efficiency, improving safety standards, and aiding in the decision-making process. Researchers have henceforth identified statistical learning methods and machine learning techniques as potent tools for excavating useful insights from this copious amount of data available. This research evaluates how these algorithms work by focusing exclusively on the analysis of sensory data collected from vessels within the maritime domain. A comparison study has been conducted between statistical learning methods (which includes regression analysis, and time series analysis) vis-a-vis machine learning approaches. The major objective of this study was to determine the most effective method for detecting anomalies while simplifying marine operations and optimizing vessel behavior. The scope of the conducted analysis is restricted to the prediction of the next trajectory points. Accurate prediction of vessel positions plays a crucial role in maritime operations, enabling efficient route planning, collision avoidance, and maritime traffic management. In this article, the authors propose a combination model that combines the benefits of Linear Regression (LR) and Long Short-Term Memory (LSTM) techniques to anticipate vessel positions based on Automatic Identification System (AIS) data. The proposed model takes advantage of the interpretability of LR and the temporal dependencies collected by LSTM to capture temporal dependencies, which improve prediction accuracy and reveal the underlying links between vessel features and future positions.

Utilizing various statistical methods to model the impact of the COVID-19 pandemic on Gross domestic product

Gross Domestic Product (GDP) is one of the key macroeconomic aggregates that measures the added value produced in a country during a period. In the contemporary world, macroeconomic uncertainty, among others due to the COVID-19 pandemic and the conflict in Ukraine, and GDP prediction remain important goals in public policy making. This study aims to predict Benin's GDP through a unidimensional statistical approach and machine learning techniques. For this purpose, GDP data were collected from the Central Bank of the West African States (BCEAO) website from 1960 to 2021. The predictions are based on comparing classical statistical and machine learning methods. For the classical statistical methods, we investigated the Autoregressive Integrated Moving Average (ARIMA) and Error Trend Seasonality (ETS) forecasting models. As for the machine learning methods, the K-Nearest Neighbors (KNN) and Long Short-Term Memory (LSTM) forecasting models proved to be sound. The findings revealed that the statistical models (ARIMA and ETS) better predict Benin's GDP. However, machine learning models (KNN and LSTM) also provide a wide range of results that can be used to analyze Benin's economic growth.

Developing statistical and machine learning models for predicting CO2 solubility in live crude oils

CO2 solubility in crude oil primarily affects the viscosity and swelling capacity of oil and determines the displacement efficiency of CO2 flooding during the enhanced oil recovery process. Dissolution of CO2 into crude oil also ensures safe and permanent storage of CO2 in depleted formations via solubility trapping. This study employed statistical and machine learning algorithms to develop precise and stable predictive models for estimating CO2 solubility in crude oils with dissolved gas. Recognizing the limitations of existing experimental data on CO2 solubility for live oils (crude oils with dissolved gas), we introduced a novel approach employing statistical and machine learning techniques for developing accurate and robust predictive models. Our methodology combined regularization, k-fold cross-validation, and statistical significance tests to ensure the robustness of the models, practices not considered in the previous studies. A linear regression model with L2 regularization is developed as the base model and compared with a 2nd order polynomial model and an ensemble feed-forward neural network (FNN) model integrated with gradient boosting (GB). The incorporation of GB into the neural network capitalizes on the additive sequence of FNN models, thereby enhancing the accuracy and reliability of predictions. Moreover, an improved user-accessible and easily implementable correlation is proposed to estimate CO2 solubility in live crude oils at diverse operating conditions as a function of reservoir temperature, saturation pressure, bubble-point pressure, oil molecular weight, and specific gravity. The performances of the developed model and the correlation are compared against the literature models and correlations. The FNN model with an overall R2 score of 0.9978, average absolute relative deviation (AARD) of 1.54 %, and root mean squared error (RMSE) of 0.0083 outperformed the previously published models and correlations. The new correlation, while being relatively simpler than previously published correlations, showed similar predictive performance with an overall R2 score of 0.9817, AARD of 4.98 %, and RMSE of 0.0225. The model and the correlation can be used to estimate CO2 solubility in live crude oils with 0.67 to 0.97 specific gravity for reservoir temperatures and pressures in the range of 28 to 124 °C and 3.23 to 32.76 MPa, respectively.

Estimation of crop evapotranspiration using statistical and machine learning techniques with limited meteorological data: a case study in Udham Singh Nagar, India

Estimation of crop evapotranspiration using statistical and machine learning techniques with limited meteorological data: a case study in Udham Singh Nagar, India

Deep Learning Approaches and Biomarkers in Medical Diagnosis

Abstract: The size of the medical information system is growing gradually. Due to this, traditional data analysis for extracting helpful information for any disease has become inefficient in providing accurate real-time valid information. Traditional data mining and statistical learning techniques, which lack sufficient domain knowledge on a complicatedly colossal amount of data, need to stop adopting new advances in deep learning technologies. Thus, the use of sophisticated machines with in-depth understanding and analysis capability is vital to provide real-time information in detecting and diagnosing diseases in the healthcare system. In this paper, we study recent deep learning approaches which are capable of working on high-dimensional and multi-dimensional data. These approaches have been deployed to identify the root cause of various diseases like Cancer, Lung Diseases, Heart Diseases, Diabetes, Hepatitis, Alzheimer’s, Dengue, Parkinson’s, etc. : In this study, our key contributions are: 1: A decent overview of the deep learning techniques. 2: Several modalities in the diagnosis of various diseases. 3: Set out recent trends in various applications of deep learning in healthcare, some of which are analysis and diagnosis of medical images, precision medicine, drug discovery, predictive analysis to support clinical decisions, and sustainable public health. 4: Several deep learning approaches with their performance are described in detail. 5: Although deep learning has achieved notable performance in detecting AD, there are several limitations, especially the availability of the patient’s data for training deep learning models on a particular disease is comparatively much less than required.

Enhancing Regional Climate Downscaling through Advances in Machine Learning

Abstract Despite the sophistication of global climate models (GCMs), their coarse spatial resolution limits their ability to resolve important aspects of climate variability and change at the local scale. Both dynamical and empirical methods are used for enhancing the resolution of climate projections through downscaling, each with distinct advantages and challenges. Dynamical downscaling is physics based but comes with a large computational cost, posing a barrier for downscaling an ensemble of GCMs large enough for reliable uncertainty quantification of climate risks. In contrast, empirical downscaling, which encompasses statistical and machine learning techniques, provides a computationally efficient alternative to downscaling GCMs. Empirical downscaling algorithms can be developed to emulate the behavior of dynamical models directly, or through frameworks such as perfect prognosis in which relationships are established between large-scale atmospheric conditions and local weather variables using observational data. However, the ability of empirical downscaling algorithms to apply their learned relationships out of distribution into future climates remains uncertain, as is their ability to represent certain types of extreme events. This review covers the growing potential of machine learning methods to address these challenges, offering a thorough exploration of the current applications and training strategies that can circumvent certain issues. Additionally, we propose an evaluation framework for machine learning algorithms specific to the problem of climate downscaling as needed to improve transparency and foster trust in climate projections. Significance Statement This review offers a significant contribution to our understanding of how machine learning can offer a transformative change in climate downscaling. It serves as a guide to navigate recent advances in machine learning and how these advances can be better aligned toward inherent challenges in climate downscaling. In this review, we provide an overview of these recent advances with a critical discussion of their advantages and limitations. We also discuss opportunities to refine existing machine learning methods alongside new approaches for the generation of large ensembles of high-resolution climate projections.

Machine Learning Methods for Attack Detection in Smart Grid

Abstract: In the realm of smart grids attack detection, statistical learning poses challenges across various attack scenarios, whether measurements are obtained either online or in batch mode. This approach categorizes measurements into two groups: secure and attacked, leveraging machine learning algorithms. The suggested method offers a framework for detecting attacks, aiming to address limitations arising from the sparse nature of the problem and leveraging any available past system knowledge. Through decision- and feature-level fusion, established batch and online learning methods are employed to tackle the attack detection challenge. To uncover unobservable attacks using statistical learning techniques, the relationships between the geometric and statistical characteristics of the attack vectors within the attack scenarios and the learning algorithms are scrutinized

Practical forecasting of risk boundaries for industrial metals and critical minerals via statistical machine learning techniques

This study examines the application of machine learning in predicting price risk boundaries for industrial metals and critical minerals, emphasizing the role of statistical dependencies among their yields. Given these commodities’ pivotal role in various industries and their influence on the global economy, accurate forecasting of their price boundaries is critical. Our research employs statistical dependencies as key features to uncover meaningful correlations essential for effective forecasting, underscoring the value of analyzing rate-based statistical dependencies alongside price fluctuations. Applying explainable artificial intelligence (xAI) techniques validates the effectiveness of combining these approaches for price boundary prediction. Our findings indicate variations in effectiveness between the Pearson correlation coefficient and normalized mutual information, challenging the Pearson coefficient’s dominance in financial analysis. The study demonstrates that the relevance of each statistical dependency varies across different machine learning models, necessitating a comprehensive analytical approach incorporating information-theoretic methods. Our analysis improves the accuracy and interpretability of price predictions by pinpointing critical relationships among selected industrial metals and critical minerals.

Machine learning in small sample neuroimaging studies: Novel measures for schizophrenia analysis.

Novel features derived from imaging and artificial intelligence systems are commonly coupled to construct computer-aided diagnosis (CAD) systems that are intended as clinical support tools or for investigation of complex biological patterns. This study used sulcal patterns from structural images of the brain as the basis for classifying patients with schizophrenia from unaffected controls. Statistical, machine learning and deep learning techniques were sequentially applied as a demonstration of how a CAD system might be comprehensively evaluated in the absence of prior empirical work or extant literature to guide development, and the availability of only small sample datasets. Sulcal features of the entire cerebral cortex were derived from 58 schizophrenia patients and 56 healthy controls. No similar CAD systems has been reported that uses sulcal features from the entire cortex. We considered all the stages in a CAD system workflow: preprocessing, feature selection and extraction, and classification. The explainable AI techniques Local Interpretable Model-agnostic Explanations and SHapley Additive exPlanations were applied to detect the relevance of features to classification. At each stage, alternatives were compared in terms of their performance in the context of a small sample. Differentiating sulcal patterns were located in temporal and precentral areas, as well as the collateral fissure. We also verified the benefits of applying dimensionality reduction techniques and validation methods, such as resubstitution with upper bound correction, to optimize performance.

Abstract 2448: Field cancerization proteomic analysis provides a novel class of functional risk biomarkers for non small cell lung cancer

Abstract Lung cancer carcinogenesis is the outcome of a field of premalignant changes that occur in the bronchial tree leading to the overt emergence of the malignant lung cancer (field cancerization). A tissue or blood proteomic signature of these premalignant changes represent a novel concept of risk biomarkers that can predict the probability of future emergence of Lung Cancer. In this study, we analyzed two independent cohorts of non-small cell lung cancer (NSCLC) patients. One cohort consisted of 18 patients who donated blood before and after tumor resection surgery and matched tissue samples (UAMS Medical Center). The second set derived from a longitudinal case-control study of plasma sample collected during the physician health study that were donated at different times prior to the diagnosis of Non-Small Cell Lung Cancer (N=272). We used affinity-enriched mass spectrometry analysis and statistical learning techniques to identify a set of common markers that discriminated plasma cases and controls and were conserved in different bronchial tissue compartments including tumor and adjacent tissue. The tumor microenvironment adjacent to and surrounding cancer cells undergoes to progressive stages of carcinogenesis during field cancerization. We identified 22 proteins involved in cell adhesion, protein folding, immune response, angiogenesis, and cancer proliferation. This is a novel class of functional risk biomarkers to determine the probability of NSCLC onset. In conclusion, our findings shed light on a new class of functional biomarkers that underly tumor pathogenesis mechanisms and can provide hints for new strategies of tumor prevention therapy or lifestyle modification. Citation Format: Rayan Ibrahim Alhammad, Ngoc Bao Vuong, Weidong Zhou, Donald J Johann Jr, Barbara Birkaya, Lance Liotta, Alessandra Luchini. Field cancerization proteomic analysis provides a novel class of functional risk biomarkers for non small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2448.

Statistical machine learning techniques applied to NIR spectral data for rapid detection of sudan dye-I in turmeric powders with optimized pre-processing and wavelength selection.

The online version contains supplementary material available at 10.1007/s13197-024-05971-9.

Intelligent Resource Orchestration for 5G Edge Infrastructures

The adoption of edge infrastructure in 5G environments stands out as a transformative technology aimed at meeting the increasing demands of latency-sensitive and data-intensive applications. This research paper presents a comprehensive study on the intelligent orchestration of 5G edge computing infrastructures. The proposed Smart 5G Edge-Cloud Management Architecture, built upon an OpenNebula foundation, incorporates a ONEedge5G experimental component, which offers intelligent workload forecasting and infrastructure orchestration and automation capabilities, for optimal allocation of virtual resources across diverse edge locations. The research evaluated different forecasting models, based both on traditional statistical techniques and machine learning techniques, comparing their accuracy in CPU usage prediction for a dataset of virtual machines (VMs). Additionally, an integer linear programming formulation was proposed to solve the optimization problem of mapping VMs to physical servers in distributed edge infrastructure. Different optimization criteria such as minimizing server usage, load balancing, and reducing latency violations were considered, along with mapping constraints. Comprehensive tests and experiments were conducted to evaluate the efficacy of the proposed architecture.

Finite mixture models for multiscale dynamics in Chinese financial markets

A novel statistical machine learning technique for decomposing, analyzing, and modeling complex time series is presented in this paper. This methodology is based on unsupervised classification methods using finite mixtures, which have been widely used in factor analysis to extract essential information from multivariate static data in a parsimonious manner. Here, they are applied in a purely dynamic framework to create a new family of autoregressive models. The selection of the number of components and appropriate geometric template in these models is based on a set of information criteria that require solving a multiple objective optimization problem. To ensure the most adequate models are selected, the problem is augmented with a simulation instruction, providing a robust criterion for assignment. Experiments on real financial data sampled over the COVID-19 pandemic demonstrate the potential of this approach as a promising alternative to the well-known Box-Jenkins methodology. The proposed method offers flexibility, accuracy, and stability, making it suitable for a variety of applications in financial modeling and other fields where complex time series data analysis is required.

Artificial intelligence: reflecting on the past and looking towards the next paradigm shift

ABSTRACT Artificial intelligence (AI) has undergone major advances over the past decades, propelled by key innovations in machine learning and the availability of big data and computing power. This paper surveys the historical progress of AI from its origins in logic-based systems like the Logic Theorist to recent deep learning breakthroughs like Bidirectional Encoder Representations from Transformers (BERT), Generative Pretrained Transformer 3 (GPT-3) and Large Language Model Meta AI (LLaMA). The early rule-based systems using handcrafted expertise gave way to statistical learning techniques and neural networks trained on large datasets. Milestones like AlexNet and AlphaGo established deep learning as a dominant AI approach. Transfer learning enabled models pre-trained on diverse corpora to excel at specialised downstream tasks. The scope of AI expanded from niche applications like playing chess to multifaceted capabilities in computer vision, natural language processing and dialogue agents. However, current AI still needs to catch up to human intelligence in aspects like reasoning, creativity, and empathy. Addressing limitations around real-world knowledge, biases, and transparency remains vital for further progress and aligning AI with human values. This survey provides a comprehensive overview of the evolution of AI and documents innovations that shaped its advancement over the past six decades.

Analysis and prediction of erosion behavior of epoxy composites using statistical and machine learning techniques

This work reports on the application of different machine learning (ML) techniques and statistical methods to analyze and predict the erosion wear performance of ramie fiber-reinforced epoxy composites. These composites, fabricated by conventional hand layup technique, are subjected to high-temperature erosion wear following a test schedule based on the design-of-experiments. The experimental database is used to determine the relative significance and contribution of individual parameter on the erosion wear rate using the analysis of variance. It suggests the impact velocity as the most significant factor contributing 81.2%, followed by impingement angle (13.56%), erodent temperature (0.06%) and stand-off distance (0.05%) to the wear rate. The data generated from experimentation are further processed to predict the erosion performance of the composites with a ML approach following four different algorithms such as support vector machine, decision tree, random forest and gradient boosting machine (GBM). The feature scores for different parameters in the above models are also evaluated to ascertain the contribution of each parameter, which consequently revealed that impact velocity and impingement angle with respective mean feature scores of 0.7538 and 0.1824 have maximum influence on the erosion rate. The performances of four different ML models are compared on the basis of their coefficient of determination (R2 value). It is found that the GBM outperforms other models with a R2 value as high as 0.9543 and has thus emerged as the best-performing prediction model. The possible mechanisms causing the wear loss are identified using electron microscopy.