Relevant Features Research Articles

Exploring Learning Analytics In E-Learning: A Comprehensive Analysis of Student Characteristics and Behavior

This article aims to explore learning analytics in e-learning through a comprehensive analysis of student characteristics and behavior. E-learning has become increasingly significant in education, particularly due to the social situation influenced by the pandemic. The Learning Management System (LMS) has become a crucial tool for educators to track and record student learning in e-learning environments. Learning analytics can aid in understanding the context of students, ensuring that they receive a personalized learning experience aligned with learning objectives. However, educators often face challenges in conducting learning analytics for e-learning students, primarily due to the large number of students to analyze and limited data availability. This study seeks to provide a detailed description of e-learning students within the Open and Distance Education (ODE) system. ODE students exhibit high diversity in demographic profiles, learning behaviors, and competency backgrounds. To support this research, we utilize datasets containing student demographic profiles and learning activity data during e-learning sessions. The datasets are obtained from the academic system and LMS log data of Universitas Terbuka. The article employs Exploratory Data Analysis (EDA) and data science approaches as the foundation for predictive and prescriptive analytics of student learning outcomes. Relevant features are extracted from the dataset to build a robust predictive model. The analysis results present patterns and relationships between student characteristics, learning behaviors, and academic achievements. This research aims to provide valuable insights for the development of more effective and personalized e-learning strategies to enhance student learning outcomes in the context of distance education. Moreover, the analysis findings can serve as a basis for informed academic decision-making to improve the quality of e-learning environments.

Prediction of Living Status for Breast Cancer Patients Using Machine Learning

Due to reasons such as diet, living habits, and genes, etc., breast cancer has gradually become a common high mortality disease and is not limited to women. Many patients suffer from terrible illness or even death due to untimely treatment or incorrect prediction of tumors. This study has 500 observations in total through utilizing the combination of oversampling and undersampling. What’s more, by using advanced machine learning, the method of recursive feature elimination and random forest, this research constructs a model and select four types of protein content in the body and age as the most relevant features in predicting patients’ living status. The model has a relative high accuracy 98.052% and can provide physicians relative accurate information to give targeted treatment. It also can be utilized as a warning to remind patients of their current physical condition. Early treatment and accurate diagnosis can significantly improve patient survival rates.

A New Approach towards a User-Driven Coastal Climate Service to Enhance Climate Resilience in European Cities

Coastal climate services play a crucial role in developing customised climate information for diverse end-users and stakeholders. To build climate-resilient societies, decision-makers should be empowered through easy access to powerful tools that enable timely adaptation to future and ongoing hazards. For this reason, fit-for-purpose climate services are needed to conduct accurate historical characterisation and projections for interpretative studies on climate- and water-related risks at the local coastal scale. The EU-funded SCORE project (Smart Control of Climate Resilience in European Coastal Cities) utilises climate and marine services for the development of smart technologies that support nature-based solutions to address specific concerns, including rising sea levels, coastal erosion, and coastal flooding due to extreme weather events. As part of the SCORE project, decision-makers will be able to address climate change-related coastal effects in their own cities through novel participatory approaches (Coastal City Living Labs—CCLLs). As part of this framework, this work (i) discusses the main requirements for the identification of fit-for-purpose coastal climate services for local-scale impact studies in European coastal cities based on CCLL requests and prior knowledge and (ii) provides relevant parameters and features that fulfil the users’ needs.

Digital Phenotyping for Mood Disorders: Methodology-Oriented Pilot Feasibility Study.

In the burgeoning area of clinical digital phenotyping research, there is a dearth of literature that details methodology, including the key challenges and dilemmas in developing and implementing a successful architecture for technological infrastructure, patient engagement, longitudinal study participation, and successful reporting and analysis of diverse passive and active digital data streams. This article provides a narrative rationale for our study design in the context of the current evidence base and best practices, with an emphasis on our initial lessons learned from the implementation challenges and successes of this digital phenotyping study. We describe the design and implementation approach for a digital phenotyping pilot feasibility study with attention to synthesizing key literature and the reasoning for pragmatic adaptations in implementing a multisite study encompassing distinct geographic and population settings. This methodology was used to recruit patients as study participants with a clinician-validated diagnostic history of unipolar depression, bipolar I disorder, or bipolar II disorder, or healthy controls in 2 geographically distinct health care systems for a longitudinal digital phenotyping study of mood disorders. We describe the feasibility of a multisite digital phenotyping pilot study for patients with mood disorders in terms of passively and actively collected phenotyping data quality and enrollment of patients. Overall data quality (assessed as the amount of sensor data obtained vs expected) was high compared to that in related studies. Results were reported on the relevant demographic features of study participants, revealing recruitment properties of age (mean subgroup age ranged from 31 years in the healthy control subgroup to 38 years in the bipolar I disorder subgroup), sex (predominance of female participants, with 7/11, 64% females in the bipolar II disorder subgroup), and smartphone operating system (iOS vs Android; iOS ranged from 7/11, 64% in the bipolar II disorder subgroup to 29/32, 91% in the healthy control subgroup). We also described implementation considerations around digital phenotyping research for mood disorders and other psychiatric conditions. Digital phenotyping in affective disorders is feasible on both Android and iOS smartphones, and the resulting data quality using an open-source platform is higher than that in comparable studies. While the digital phenotyping data quality was independent of gender and race, the reported demographic features of study participants revealed important information on possible selection biases that may result from naturalistic research in this domain. We believe that the methodology described will be readily reproducible and generalizable to other study settings and patient populations given our data on deployment at 2 unique sites.

Curating Multimodal Satellite Imagery for Precision Agriculture Datasets with Google Earth Engine

In the era of modern agriculture, satellite imagery has been widely used to monitor crops, one of which is paddy. This paper tries to describe the vegetation indices, climate, and soil index features related to paddy plants and curates a collection of satellite imagery on the Google Earth Engine (GEE). This paper reveals how GEE can be used to collect and process multimodal satellite imagery to form a precision agriculture dataset. The objective of this study is to establish a comprehensive precision agriculture dataset by leveraging multimodal satellite imagery to monitor paddy crops. The data collected as a dataset originates from 306 locations in Karawang Regency, Indonesia, during the 2019-2020 period. In the first step, we identify the relevant features essential for paddy crop analysis. Subsequently, we carefully select image collections within GEE based on these features. Afterward, we perform data acquisition and necessary preprocessing through the Google Colab environment. The results showed that satellite imagery from Sentinel-2 outperforms Landsat 8 in terms of spatial and temporal resolution. Apart from that, the generated dataset successfully captures the growth patterns of paddy plants.

Rapid screening of dyes for self-aggregation, adsorption and metabolic integrity – Quantitative metrics as a prelude to biological studies

The traditional development of dyes for biological applications typically begins with synthesis, is followed by spectroscopic/photophysical characterization, and culminates in testing with cultured cells or in animals. The present work describes a set of facile, low-technology, invitro, largely quantitative absorption spectrophotometric assays in aqueous solution as a prelude to biological studies. The assays include the following: (1) Self-aggregation in phosphate-buffered saline (PBS), with or without the presence of 3 % bovine serum albumin (BSA), via reciprocal change in concentration and pathlength; (2) Adsorption to solid materials as biomolecular surrogates, including cellulose, chitosan, polyethylene (PE), styrene-divinylbenzene beads (XAD-4), polyacrylic beads (XAD-7), anion exchange beads (IRA-67), cation exchange beads (CG-50), polytetrafluoroethylene (PTFE) membranes, and Nylon membranes; (3) Hydrophilicity/hydrophobicity assessment, by measurement (not calculation) of logP values; and (4) Metabolic integrity, by exposure to a rat liver homogenate. Twenty-two dyes were examined in the assays. The dyes represent diverse classes (acridine, anthraquinone, azo, benzimidazole, coumarin, cyanine, flavonoid, naphthalene, phthalocyanine, polyene, tetrapyrrole, triarylmethine, xanthene), polarity (anionic, cationic, nonionic, hydrophobic, amphipathic), and absorption (ultraviolet, visible, near-infrared). As one example, indocyanine green (ICG), an FDA-approved dye for invivo use, (i) aggregates at the concentration of 5 μM but disaggregates in the presence of BSA; (ii) in PBS is very adsorptive (>80 % binding) to various materials (cellulose, chitosan, PE, IRA-67, CG-50, PTFE, Nylon) and quite adsorptive (50–80 % binding) to XAD-4 and XAD-7, (iii) exhibits only marginal preference for water versus 1-octanol (1.5:1), and (iv) is largely digested by rat liver homogenate. A single dye can be examined in the span of one day. The quantitative metrics should prove valuable for establishing biologically relevant features of diverse dyes by simple studies in aqueous solution.

Twin-SegNet: Dynamically coupled complementary segmentation networks for generalized medical image segmentation

Image segmentation is deemed an important task in biomedicine, often required for proper diagnosis and prognosis of many diseases. Deep learning (DL) based segmentation methods have received considerable attention in recent years due to the increasing availability of clinical datasets. Many novel ideas have been proposed over the years driving progress in the field of automatic segmentation research. Contrary to the theme of contemporary literature, we demonstrate that considering the background tissue segmentation task alongside the main foreground task can improve overall segmentation performance when considered from a general medical image segmentation perspective. We, therefore, propose a DL framework called Twin Segmentation Network (Twin-SegNet) that ties together two streams (foreground and background) through an image reconstruction task. A boxed Mean Squared Error loss is proposed to complement the dice losses from both streams. We furthermore propose a Wavelet Convolutional Block (WCB) to enhance the edge information extracting capabilities of both streams and also a Partial Channel Recalibration (PCR) block to allow mutual feature exchange between the two streams so that each stream can emphasize more on channels with more discriminative and relevant features. We present experimental results on five public datasets: BUSI, GLAS, ISIC-2018, MoNuSeg, and CVC-ClinicDB. Unlike conventional baselines that demonstrate convincing performance in some datasets and poor performance in others, Twin-SegNet is able to consistently achieve state-of-the-art results with impressive F1 scores of 88.46%, 93.11%, 91.61%, 81.78% and 94.69% on each dataset respectively, showing its great potential as a general segmentation framework.

Supervised Feature Selection based on the Law of Total Variance

Feature selection is a fundamental pre-processing step in machine learning that decreases data dimensionality by removing superfluous and irrelevant features. This study proposes a supervised feature selection method based on feature relevance by employing the law of total variance (LTV). Specifically, the LTV is used to quantify the relevance of features by analysing the association between features and class label. Six classifiers were employed to evaluate the performance and reliability of the proposed method pertaining to classification accuracy. The results proved that a feature subset given by the proposed method has the capability to achieve comparable classification accuracy to the full feature set when just half or less than half of the original features are retained. The proposed method was also proven to be versatile as it can achieves adequate classification accuracy with all six classifiers with different learning schemes. In addition, a comparison with a similar type of feature selection method (AmRMR) shows that the proposed method yields a more accurate classification.

MYTHOLOGICAL PERSPECTIVE OF THE LINGUISTIC WORLDVIEW (THE MOON AND THE SUN IN PHRASEOLOGY)

In the focus of the article there is the problem of hidden mythological seman- tics of lexical (more broadly — phraseological) units in the Slavic and Germanic (English) languages. Often the lack of proper attention to this problem leads to inaccurate, inappropriate, superficial translation or to incorrect assessment of certain fragments of folklore texts of various genres in their ethnolinguistic study. Being «erased» on the surface of linguistic consciousness, which in the course of evolution has developed new stereotypes, the structures of mythological think- ing are often preserved at the level of lexical semantics and in the semantics of phraseological units. The mythological plane of the semantics of linguistic units is an important level of such a complex object of linguistics as the linguistic picture of the world, since it largely determines the specificity and originality of linguistic thinking, while remaining implicit, hidden behind the everyday, habitual meaning. On the example of microtexts (established figurative phrases) with two oppo- site ancient concepts «the moon» and «the sun», the commonality and difference in the mythological semantics of the units of the Slavic and English languages are shown. The relevant differential features of the Slavic and Germanic worldviews are emphasized.

An automated detection and classification of brain tumor from MRIs using Water Chaotic Fruitfly Optimization (WChFO) based Deep Recurrent Neural Network (DRNN)

AbstractDue to the complexity of the images and dearth of anatomical models, it is highly difficult to accurately represent the various deformations in each component of the medical images. In recent years, a significant number of children and adults have affected from brain tumors, which is one of the most terrible types of disease affects the people around the world. Moreover, the Magnetic Resonance Imaging (MRI) based brain tumor detection is one of a significant study area in the field of medical imaging. Since, the use of computerized methods aids in the detection and treatment of disease by the medical professionals. The development of an automated method for the accurate detection and classification of tumors from brain MRIs. In this framework, a tanh normalization process is used to smooth out the input brain MRIs with less noise artefacts and improved quality. Then, a group feature extraction model is used to extract the relevant features from the normalized image, which includes both Speeded Up Robust Features (SURF) and Grey Level Co-occurrence Matrix (GLCM) features. The Water Chaotic Fruitfly Optimization (WChFO) method is used to identify the best features for increasing the speed of classifier training and testing processes with less time. Moreover, a Deep Recurrent Neural Network (DRNN) model is used to classify the type of brain tumor for accurate early diagnosis and treatment. The most well-known benchmarking datasets, like BRATS and Kaggle, employed for analysis in order to assess the effectiveness and results of the proposed brain tumor diagnosis system. By using the proposed WChFO-DRNN technique, the accuracy of the tumor detection system is increased to 99.2% with the sensitivity, specificity of 99% and time consumption of 0.2s.

WaQuPs: A ROS-Integrated Ensemble Learning Model for Precise Water Quality Prediction

Water presents challenges in swiftly and accurately assessing its quality due to its intricate composition, diverse sources, and the emergence of new pollutants. Current research tends to oversimplify water quality, categorizing it as potable or not, despite its complexity. To address this, we developed a water quality prediction system (WaQuPs), a sophisticated solution tackling the intricacies of water quality assessment. WaQuPs employs advanced machine learning, including an ensemble learning model, categorizing water quality into nuanced levels: potable, lightly polluted, moderately polluted, and heavily polluted. To ensure rapid and precise dissemination of information, WaQuPs integrates an Internet of Things (IoT)-based communication protocol for the efficient delivery of detected water quality results. In its development, we utilized advanced techniques, such as random oversampling (ROS) for dataset balance. We used a correlation coefficient to select relevant features for the ensemble learning algorithm based on the Random Forest algorithm. Further enhancements were made through hyperparameter tuning to improve the prediction accuracy. WaQuPs exhibited impressive metrics, achieving an accuracy of 83%, precision of 82%, recall of 83%, and an F1-score of 82%. Comparative analysis revealed that WaQuPs with the Random Forest model outperformed both the XGBoost and CatBoost models, confirming its superiority in predicting water quality.

Exploring the role of interest rates, macroeconomic environment, agricultural cycle, and gender on loan demand in the agricultural sector: Evidence from Mali

AbstractFormal credit plays an important role for the development of the agriculture sector in developing countries because many farmers are characterized as liquidity constrained. Access to credit can increase farmers' purchasing power for inputs and agricultural technology, thus raising the overall productivity. Farmers in Mali are particularly vulnerable to shocks, such as heavy precipitation events. Access to liquidity to increase the resilience of the agricultural sector is essential. Therefore, higher financing volumes are required, which make the analysis of loan demand in agriculture of interest. The purpose of this paper is to empirically investigate the role of the interest rate, the macroeconomic environment, the agricultural cycle and the gender of the farmer on the loan demand in the agricultural sector from a country in the Sahel. Unique and comprehensive loan data at the farm level, provided by a commercial Malian bank, is used for this analysis. The analysis covers the period from 2010 to 2020. Two different estimation strategies are combined. First, an ordinary least square regression is applied with the granted loan amount as the dependent variable. Second, the machine learning technique, least absolute shrinkage and selection operator, is applied to select the most relevant features to be used as explanatory variables in the estimation. The results reveal that the interest rate, the gross value added, the farmer's gender as well as the agricultural cycle have statistically significant effects on the granted loan demand in agriculture. These results are of interest to policymakers, who deal with financial inclusion as well as market failures, and agricultural financial institutions who could incorporate such information in the design of future loan products to stimulate farmers' loan demand, especially for female farmers. [EconLit Citations: G20, G21, O13, O16, Q14, Q18].

Levels of Variation in Subordinates of Immediate Succession in Current Spanish

In this paper, I analyze, from a compositional perspective, the relevant features to construct the interpretation of immediate succession between a subordinate event and the event that takes place in the main sentence. Among all the components involved in the construction of the meaning of immediate succession, I focus particularly on the subordinators, which present a mosaic of variation in current Spanish. The key ideas that can be derived from the data analysis are the following. First: subordinators of immediate succession are the loci of variation of temporal subordinates. Second: a subordinator of immediate succession is a “linguistic variable” that can be syntactically materialized in different forms by applying general rules that do not change the meaning, although sometimes they do change the grammatical category. Third: in the diachronic evolution of Spanish, several patterns of internal structure have emerged for immediate succession subordinators. However, most of them have ceased to be productive, although some subordinators that were coined with these patterns have survived as fossils in the current language. Fourth: the only productive pattern in the present language can be reduced to the Adv (immediacy) + que scheme, which goes back to Late Latin.

Porous polymer monolithic columns to investigate the interaction of humic substances with herbicides and emerging pollutants by affinity chromatography

BackgroundUnderstanding the interaction mechanisms and the relevant binding constants between humic acids and emerging or regulated pollutants is of utmost importance in predicting their geochemical mobility, bioavailability, and degradation. Fluorescence spectroscopy, UV–vis spectroscopy, equilibrium dialysis, and solid-phase extraction combined with liquid chromatography-mass spectrometry have been employed to elucidate interactions of humic acids with organic micropollutants, especially pharmaceutical drugs. These methods demand large sample volumes, long equilibration times, and laborious extraction steps which may imply analytical errors. Monolithic high-performance affinity chromatography is an alternative and simpler method to investigate these interactions and determine the binding constants. ResultsPolymer monoliths based on aminated glycidyl methacrylate and ethylene glycol dimethacrylate served to immobilize Cu(II) and then humic acid to produce monolithic affinity chromatography columns with humic acid as the active interaction phase. About 86.5 mg of humic acid was immobilized per gram of polymer. The columns enabled a comparison of the binding strength of humic acid with herbicides and emerging pollutants at 25 °C and pH 6.0 ± 0.1. Paracetamol, acetylsalicylic acid, and salicylic acid did not retain. Among the compounds that interacted with humic acid, the order of increasing affinity, estimated by the global affinity constant (nKa) or partition coefficient (KD) was: caffeine < simazine < atrazine ∼ propazine < benzophenone. The nKa (L mol−1) values ranged from (4.9 ± 0.3) × 102 for caffeine to (1.9 ± 0.3) × 103 for benzophenone, whereas KD (L kg−1) varied from 14 ± 1 to 56 ± 8 for the same compounds. Significance and noveltyTo our knowledge, this is the first paper demonstrating the use of a monolithic platform to immobilize supramolecular structures of humic acids exploiting immobilized metal affinity to comparatively evaluate their affinity towards emerging pollutants exploiting the concepts of high-performance affinity chromatography. The proposed approach needs only small amounts of humic acid, which is a relevant feature in preparing columns with humic substances isolated and purified from remote areas.

Machine learning-based diagnosis in wave power plants for cost reduction using real measured experimental data: Mutriku Wave Power Plant

In comparison to wind farms, the relative scarcity of actual operational data from wave power plants has contributed to a significant research gap in the areas of wave farm forecasting and cost reduction. In this context, this manuscript presents a new Machine Learning-based Power Take-Off (PTO) diagnosis for wave energy generation farms which has the potential to serve as an extensive reference for other wave energy farms and offer substantial benefits to both investors and policymakers involved in the advancement of the emerging wave technologies. The suggested method has been employed at the Mutriku Wave Power Plant (WWP) to facilitate the implementation of predictive maintenance strategies and reduce the Levelized Cost of Energy (LCoE). Hence, the research study considers two main extraction methods, namely, Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA), used to select the most relevant features for OWC diagnosis. In addition, two classification methods have been considered: Support Vector Machine (SVM) and Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN). The obtained data show that, although both methods allow to achieve an effective performance with an excellent degree of accuracy, the ANN-based method presents better results with 98% accuracy against 81% for the SVM when using PCA extraction method. Then, the developed classification-based OWC diagnosis has been used for the development of a predictive maintenance strategy at the Mutriku WPP, analyzing its impact on the economic indicators. The results indicate that, using the proposed predictive maintenance strategy, the OpEx may be decreased down to 17%, downtime may be decreased down to 55% and plant availability may be better up to 95%, leading to a 5% LCoE reduction.

Improving wave height prediction accuracy with deep learning

A novel convolutional neural network-long short-term memory (CNN-LSTM) model is proposed for wave height prediction. The model effectively extracts relevant features such as wind speed, wind direction, wave height, latitude, and longitude. The proposed model outperforms traditional machine learning algorithms such as multi-layer perceptron (MLP), support vector machine (SVM), random forest and LSTM, especially for extreme values and fluctuations. The model has a significantly lower average root mean square error (RMSE) of 71.1%, 72.8%, 71.9% and 72.2% for MLP, SVM, random forest and LSTM, respectively. Our model is computationally more efficient than traditional numerical simulations, making it suitable for real-time applications. Moreover, it has better long-term robustness compared to traditional models. The integration of CNN and LSTM techniques improves wave height prediction accuracy while enhancing its efficiency and robustness. The proposed CNN-LSTM model provides a promising tool for effective wave height prediction, making a valuable contribution to coastal disaster prevention and mitigation. Future research should aim to improve long-term prediction accuracy, and we believe that the CNN-LSTM model plays a crucial role in developing real-time coastal disaster prevention and mitigation measures. Overall, our study represents a significant step towards achieving more accurate and efficient wave height prediction using machine learning techniques.

Analysis and selection of relevant network anomaly detection metrics

The object of the study is the detection of network anomalies - an important and dynamically developing area of research. The article discusses the main aspects of network anomaly detection. Principles are formulated that allow generalization of various anomaly detection methods. The attacks that network intrusion detection systems typically face are presented, along with the characteristics and types of intrusion detection methods. Network anomalies are considered as manifestations of network attacks, which makes it possible to classify anomalies. The analysis of iterative sources showed that, despite the breadth of coverage of various methods, subject areas and tasks for detecting network anomalies, less attention is paid to the key issue - the analysis of network anomaly metrics and the rationale for choosing the relevant metric in a particular case. The paper presents types, characteristics and examples of network anomalies. To classify and facilitate the detection of network anomalies, metrics are proposed that are based on proximity measures for numeric, categorical, and mixed data types that characterize anomalies. The network anomaly detection problem is presented as a classification or clustering problem. The components that characterize this problem are identified, namely types of input data, acceptability of proximity measures, data labeling, classification of methods based on the use of labeled data, identifying relevant features and reporting anomalies. An approach is described that allows you to timely generate the required set of metrics, which will ensure not only the formation of preventive countermeasures, but also allow you to assess the current state of the security system as a whole. In addition, it provides the possibility of forming multi-circuit security systems, taking into account the influence (integration) of targeted (mixed) attacks on infrastructure elements, as well as the possibility of their synthesis with social engineering methods.

VaBTFER: An Effective Variant Binary Transformer for Facial Expression Recognition.

Existing Transformer-based models have achieved impressive success in facial expression recognition (FER) by modeling the long-range relationships among facial muscle movements. However, the size of pure Transformer-based models tends to be in the million-parameter level, which poses a challenge for deploying these models. Moreover, the lack of inductive bias in Transformer usually leads to the difficulty of training from scratch on limited FER datasets. To address these problems, we propose an effective and lightweight variant Transformer for FER called VaTFER. In VaTFER, we firstly construct action unit (AU) tokens by utilizing action unit-based regions and their histogram of oriented gradient (HOG) features. Then, we present a novel spatial-channel feature relevance Transformer (SCFRT) module, which incorporates multilayer channel reduction self-attention (MLCRSA) and a dynamic learnable information extraction (DLIE) mechanism. MLCRSA is utilized to model long-range dependencies among all tokens and decrease the number of parameters. DLIE's goal is to alleviate the lack of inductive bias and improve the learning ability of the model. Furthermore, we use an excitation module to replace the vanilla multilayer perception (MLP) for accurate prediction. To further reduce computing and memory resources, we introduce a binary quantization mechanism, formulating a novel lightweight Transformer model called variant binary Transformer for FER (VaBTFER). We conduct extensive experiments on several commonly used facial expression datasets, and the results attest to the effectiveness of our methods.

A novel hypoglycemia alarm framework for type 2 diabetes with high glycemic variability.

In patients with type 2 diabetes (T2D), accurate prediction of hypoglycemic events is crucial for maintaining glycemic control and reducing their frequency. However, individuals with high blood glucose variability experience significant fluctuations over time, posing a challenge for early warning models that rely on static features. This article proposes a novel hypoglycemia early alarm framework based on dynamic feature selection. The framework incorporates domain knowledge and introduces multi-scale blood glucose features, including predicted values, essential for early warnings. To address the complexity of the feature matrix, a dynamic feature selection mechanism (Relief-SVM-RFE) is designed to effectively eliminate redundancy. Furthermore, the framework employs online updates for the random forest model, enhancing the learning of more relevant features. The effectiveness of the framework was evaluated using a clinical dataset. For T2D patients with a high coefficient of variation (CV), the framework achieved a sensitivity of 81.15% and specificity of 98.14%, accurately predicting most hypoglycemic events. Notably, the proposed method outperformed other existing approaches. These results indicate the feasibility of anticipating hypoglycemic events in T2D patients with high CV using this innovative framework.

A powerful Peripheral Arterial Disease detection using machine learning-based severity level classification model and hyper parameter optimization methods

This study addresses the critical need for improved detection and assessment of Peripheral Arterial Disease (PAD) using Artificial Intelligence (AI) algorithms, aiming to reduce global mortality rates associated with this condition. The research focuses on enhancing the efficiency of machine learning (ML) models for PAD detection and severity categorization by employing optimization techniques. Notably, this study introduces the application of Synthetic Minority Oversampling Technique (SMOTE) to mitigate imbalanced data challenges in PAD detection. The primary objective of this research is to develop highly efficient ML models utilizing six distinct AI classifiers and Hyper Parameter Optimization (HPO), specifically targeting PAD identification and severity level categorization. Leveraging the Cleveland and Statlog datasets, the study constructs and evaluates these models using all relevant features. The principal findings indicate remarkable advancements in PAD detection accuracy. The proposed models, optimized using SMOTE and Enhanced Decision Tree (EDT) through HPO, surpass state-of-the-art methodologies, achieving exceptional accuracies of 99.20% and 98.52% on the Cleveland and Statlog datasets, respectively. These models outperform existing approaches, assessed using comprehensive metrics such as F1 score, Accuracy, Precision, Recall, Specificity, and MCC. In conclusion, this study's innovative approach significantly improves the capability to assess PAD and categorize its severity levels accurately. The proposed models hold promise in aiding physicians to evaluate patients' cardiac conditions effectively, enabling early interventions that have the potential to mitigate mortality rates associated with cardiovascular diseases.