Multiple Performance Metrics Research Articles

Neural Inference Search for Multiloss Segmentation Models.

Semantic segmentation is vital for many emerging surveillance applications, but current models cannot be relied upon to meet the required tolerance, particularly in complex tasks that involve multiple classes and varied environments. To improve performance, we propose a novel algorithm, neural inference search (NIS), for hyperparameter optimization pertaining to established deep learning segmentation models in conjunction with a new multiloss function. It incorporates three novel search behaviors, i.e., Maximized Standard Deviation Velocity Prediction, Local Best Velocity Prediction, and n -dimensional Whirlpool Search. The first two behaviors are exploratory, leveraging long short-term memory (LSTM)-convolutional neural network (CNN)-based velocity predictions, while the third employs n -dimensional matrix rotation for local exploitation. A scheduling mechanism is also introduced in NIS to manage the contributions of these three novel search behaviors in stages. NIS optimizes learning and multiloss parameters simultaneously. Compared with state-of-the-art segmentation methods and those optimized with other well-known search algorithms, NIS-optimized models show significant improvements across multiple performance metrics on five segmentation datasets. NIS also reliably yields better solutions as compared with a variety of search methods for solving numerical benchmark functions.

Machine Learning Model of Emergency Department Use for Patients Undergoing Treatment for Head and Neck Cancer Using Comprehensive Multifactor Electronic Health Records

To use a hybrid method, combining statistical profiling, machine learning (ML), and clinical evaluation to predict emergency department (ED) visits among patients with head and neck cancer undergoing radiotherapy. Patients with head and neck cancer treated with radiation therapy from 2015 to 2019 were identified using electronic health record data. Records from 60 days before 90 days after treatment were analyzed. Statistical profiling and ML were used to create a predictive model for ED visits during or after radiation therapy. A comprehensive set of variables were studied. Multiple ML models were developed including extreme gradient-boosted decision tree and generalized logistic regression with comparison of multiple predictive performance metrics. Of the 1,355 patients studied, 13% had an ED visit during or after treatment. Our hybrid methodology enabled evidence-based winnowing of candidate features from 141 to 11 with clinically applicable, evidence-based thresholds. Extreme gradient boosting had the highest area under the curve (0.81 ± 0.06) with a sensitivity of 0.89 ± 0.10 and exceeded generalized logistic regression (area under the curve 0.64 ± 0.02). Significant predictors of ED visits during treatment included increasingly complex opioid use, number of prior ED visits, tumor volume, rate of change of blood urea nitrogen, total bilirubin, body mass index, and distance from hospital. Our approach combining bootstrapped statistical profiling and ML importance analysis supported integration of clinician input to identify a distilled set of phenotypical characteristics for developing ML models predicting which patients undergoing head and neck cancer radiation therapy were at risk for ED visits.

An Energy-Efficient Hybrid Clustering Technique (EEHCT) for IoT-Based Multilevel Heterogeneous Wireless Sensor Networks

Internet-of-Things (IoT)-based Heterogeneous Wireless Sensor Network (HWSN) has emerged as a prevalent technology that plays a significant role in developing various human-centric applications. Like in a wireless sensor network (WSN), energy is also the most crucial resource in IoT-based HWSN. The researchers have proposed many works to achieve energy-efficient network operations by minimizing energy usage. A vast proportion of these works emphasize using the clustering approach, which has proved its worth to a great extent. However, most schemes require the repeated formation of clusters incurring a significant amount of nodes’ energy in the clustering process. The protocol design of such schemes also varies with the changing levels of heterogeneity. In this work, a hybrid clustering scheme- An Energy-Efficient Hybrid Clustering Technique (EEHCT) has been proposed for IoT-based HWSN that minimizes the energy consumption in clusters’ formation and distributes the network load evenly irrespective of the heterogeneity level to prolong network lifetime. It appropriately utilizes dynamic and static clustering strategies to formulate the load-balanced clusters in the network. EEHCT establishes its supremacy over state-of-the-art schemes via an extensive set of simulations and experimentation in terms of multiple network performance metrics like stability, throughput, and network lifetime. Like, it achieves a gain up to 90.27% with respect to network lifetime over its peers in the standard operating conditions and under varying network configurations. In addition to quantitative analysis, a statistical analysis has also been provided to demonstrate the formation of energy-balanced clusters through the proposed scheme.

Learning EKG Diagnostic Models with Hierarchical Class Label Dependencies.

Electrocardiogram (EKG/ECG) is a key diagnostic tool to assess patient's cardiac condition and is widely used in clinical applications such as patient monitoring, surgery support, and heart medicine research. With recent advances in machine learning (ML) technology there has been a growing interest in the development of models supporting automatic EKG interpretation and diagnosis based on past EKG data. The problem can be modeled as multi-label classification (MLC), where the objective is to learn a function that maps each EKG reading to a vector of diagnostic class labels reflecting the underlying patient condition at different levels of abstraction. In this paper, we propose and investigate an ML model that considers class-label dependency embedded in the hierarchical organization of EKG diagnoses to improve the EKG classification performance. Our model first transforms the EKG signals into a low-dimensional vector, and after that uses the vector to predict different class labels with the help of the conditional tree structured Bayesian network (CTBN) that is able to capture hierarchical dependencies among class variables. We evaluate our model on the publicly available PTB-XL dataset. Our experiments demonstrate that modeling of hierarchical dependencies among class variables improves the diagnostic model performance under multiple classification performance metrics as compared to classification models that predict each class label independently.

A Hybrid System of Braden Scale and Machine Learning to Predict Hospital-Acquired Pressure Injuries (Bedsores): A Retrospective Observational Cohort Study

Background: The Braden Scale is commonly used to determine Hospital-Acquired Pressure Injuries (HAPI). However, the volume of patients who are identified as being at risk stretches already limited resources, and caregivers are limited by the number of factors that can reasonably assess during patient care. In the last decade, machine learning techniques have been used to predict HAPI by utilizing related risk factors. Nevertheless, none of these studies consider the change in patient status from admission until discharge. Objectives: To develop an integrated system of Braden and machine learning to predict HAPI and assist with resource allocation for early interventions. The proposed approach captures the change in patients' risk by assessing factors three times across hospitalization. Design: Retrospective observational cohort study. Setting(s): This research was conducted at ChristianaCare hospital in Delaware, United States. Participants: Patients discharged between May 2020 and February 2022. Patients with HAPI were identified from Nursing documents (N = 15,889). Methods: Support Vector Machine (SVM) was adopted to predict patients' risk for developing HAPI using multiple risk factors in addition to Braden. Multiple performance metrics were used to compare the results of the integrated system versus Braden alone. Results: The HAPI rate is 3%. The integrated system achieved better sensitivity (74.29 ± 1.23) and detection prevalence (24.27 ± 0.16) than the Braden scale alone (sensitivity (66.90 ± 4.66) and detection prevalence (41.96 ± 1.35)). The most important risk factors to predict HAPI were Braden sub-factors, overall Braden, visiting ICU during hospitalization, and Glasgow coma score. Conclusions: The integrated system which combines SVM with Braden offers better performance than Braden and reduces the number of patients identified as at-risk. Furthermore, it allows for better allocation of resources to high-risk patients. It will result in cost savings and better utilization of resources. Relevance to clinical practice: The developed model provides an automated system to predict HAPI patients in real time and allows for ongoing intervention for patients identified as at-risk. Moreover, the integrated system is used to determine the number of nurses needed for early interventions. Reporting Method: EQUATOR guidelines (TRIPOD) were adopted in this research to develop the prediction model. Patient or Public Contribution: This research was based on a secondary analysis of patients' Electronic Health Records. The dataset was de-identified and patient identifiers were removed before processing and modeling.

Reinforcement learning-driven deep question generation with rich semantics

Deep question generation (DQG) refers to generating a complex question from different sentences in context. Existing methods mainly focus on enhancing information extraction based on the encoder–decoder neural networks though they cannot perform well in DQG tasks. To address this issue, we consider combining reinforcement learning with semantic-rich information to generate deep questions in this paper. In particular, we propose a Semantic-Rich Reinforcement Learning Deep Question Generation (SRL-DQG) model, which better utilizes the semantic graphs of document representations based on the Gated Graph Neural Network (GGNN). In order to generate high-quality questions, we also optimize specific objectives via reinforcement learning with consideration of four evaluation factors including naturality, relevance, answerability, and difficulty. Empirical evaluations demonstrate that our SRL-DQG effectively improves the quality of generated questions and achieves superior performance than existing methods in terms of multiple performance metrics. Specifically, we show that several BLEU-n scores were improved by 3.5% to 10% after running SRL-DQG on 6072 samples of HotPotQA.

Evolutionary Computing to solve product inconsistencies in Software Product Lines

• Inconsistencies prevent and delay the configuration of an SPL product. • Evolutionary computation is proposed to resolve inconsistencies in SPL products. • ga-SPLIS (genetic algorithm-Software Product LIne Solution) is developed. • Results show the applicability of GA to resolve the SPL product inconsistencies. In Software Product Lines (SPLs), multiple design teams work collectively to configure products. Often, having multiple sub-designs leads to inconsistencies which prevent and delay the configuration of the final product at the end. Solutions are available to identify and resolve these inconsistencies for limited scale SPLs. For large scale and industrial SPLs, resolution of inconsistencies is still a potential research topic. In this paper, we propose, implement and validate an application of evolutionary computation, particularly genetic algorithms, to resolve inconsistencies in large scale SPL products. We label our approach ga -SPLIS (genetic algorithm-Software Product LIne Solution). Using a concrete experimental methodology with different feature set sizes and multiple standard performance metrics, we generate results which validate the applicability of ga -SPLIS for the SPL community.

Freeway Traffic Speed Prediction under the Intelligent Driving Environment: A Deep Learning Approach

The intelligent transportation system (ITS) has been proven capable of effectively addressing traffic congestion issues. For vehicles to perform effectively and improve mobility under the intelligent driving environment, real-time prediction of traffic speed is undoubtedly essential. Considering the complex spatiotemporal dependency inherent in traffic data, conventional prediction models encounter many limitations. To improve the prediction performance and investigate the temporal features, this study focuses on emerging deep neural networks (DNNs) using the Caltrans Performance Measurement System (PeMS) data. This research also establishes an intelligent driving environment in the simulation and compares the traditional car-following model with deep learning methods in terms of multiple performance metrics. The results indicate that both supervised learning and unsupervised learning are superior to the simulation-based model on the freeway, and the two deep learning networks are almost identical to one another. Besides, the result reveals that all models have their latent features for different time dimensions under the low traffic loads, transition states, and heavy traffic loads. This is critical in the application of prediction technologies in ITS. The findings can assist transportation researchers and traffic engineers in both traffic operation and management, such as bottleneck identification, platooning control, and route planning.

Functional genomics of OCTN2 variants informs protein-specific variant effect predictor for Carnitine Transporter Deficiency

Genetic variants in SLC22A5, encoding the membrane carnitine transporter OCTN2, cause the rare metabolic disorder Carnitine Transporter Deficiency (CTD). CTD is potentially lethal but actionable if detected early, with confirmatory diagnosis involving sequencing of SLC22A5. Interpretation of missense variants of uncertain significance (VUSs) is a major challenge. In this study, we sought to characterize the largest set to date (n = 150) of OCTN2 variants identified in diverse ancestral populations, with the goals of furthering our understanding of the mechanisms leading to OCTN2 loss-of-function (LOF) and creating a protein-specific variant effect prediction model for OCTN2 function. Uptake assays with 14C-carnitine revealed that 105 variants (70%) significantly reduced transport of carnitine compared to wild-type OCTN2, and 37 variants (25%) severely reduced function to less than 20%. All ancestral populations harbored LOF variants; 62% of green fluorescent protein (GFP)-tagged variants impaired OCTN2 localization to the plasma membrane of human embryonic kidney (HEK293T) cells, and subcellular localization significantly associated with function, revealing a major LOF mechanism of interest for CTD. With these data, we trained a model to classify variants as functional (>20% function) or LOF (<20% function). Our model outperformed existing state-of-the-art methods as evaluated by multiple performance metrics, with mean area under the receiver operating characteristic curve (AUROC) of 0.895 ± 0.025. In summary, in this study we generated a rich dataset of OCTN2 variant function and localization, revealed important disease-causing mechanisms, and improved upon machine learning-based prediction of OCTN2 variant function to aid in variant interpretation in the diagnosis and treatment of CTD.

A comparative study for landslide susceptibility assessment using machine learning algorithms based on grid unit and slope unit

Landslide susceptibility assessment is an important support for disaster identification and risk management. This study aims to analyze the application ability of machine learning hybrid models in different evaluation units. Three typical machine learning models, including random forest forest by penalizing attributes (FPA) and rotation forest were merged by random subspace algorithm. Twelve evaluation factors, including elevation, slope angle, slope aspect, roughness, rainfall, lithology, distance to rivers, distance to roads, normalized difference vegetation index, topographic wetness index, plan curvature, and profile curvature, were extracted from 155 landslides in Yaozhou District, Tongchuan City, China. Six landslide susceptibility maps were generated based on the slope units divided by curvature and 30 m resolution grid units. Multiple performance metrics showed that the RS-RF model based on slope units has excellent spatial prediction ability. At the same time, the method of slope unit division based on curvature is proved to be more suitable for the typical Loess tableland regions, which provides basis for the selection of evaluation units in landslide susceptibility assessment.

Research on the Ability to Detect Fake News with Machine Learning

The task of classifying news manually requires in-depth knowledge of the domain and expertise to spot anomalies in the text. During this research, we discussed the matter of classifying fake news articles using machine learning models and ensemble techniques. The info we used in our work is collected from the World Wide Web and contains news articles from various domains to cover most of the news rather than specifically classifying political news. The first aim of the research is to identify patterns in text that differentiate fake articles from true news. Within the proposed system we will extract different textual features from the articles using a machine learning tool and used the feature set as an input to the models. the training models were trained and parameter-tuned to obtain optimal accuracy. Some models have achieved relatively higher correctness than others. we'll use multiple performance metrics to compare the results for each algorithm. The ensemble learners have shown an whole better score on all presentation metrics as related to the separable beginners.

Multiobjective Inverse Design of Solid-State Quantum Emitter Single-Photon Sources

Single solid-state quantum emitters offer considerable potential for the implementation of sources of single indistinguishable photons, which are central to many photonic quantum information systems. Nanophotonic geometry optimization with multiple performance metrics is imperative to convert a bare quantum emitter into a single-photon source that approaches the necessary levels of purity, indistinguishability, and brightness for quantum photonics. We present an inverse design methodology that simultaneously targets two important figures-of-merit for high-performance quantum light sources: the Purcell radiative rate enhancement and the coupling efficiency into a desired light collection channel. We explicitly address geometry-dependent power emission, a critical but often overlooked aspect of gradient-based optimization of quantum emitter single-photon sources. We illustrate the efficacy of our method through the design of a single-photon source based on a quantum emitter in a GaAs nanophotonic structure that provides a Purcell factor Fp = 21 with a 94% waveguide coupling efficiency, while respecting a geometric constraint to minimize emitter decoherence caused by etched sidewalls. Our results indicate that multiobjective inverse design can yield competitive performance with more favorable trade-offs than conventional approaches based on a pre-established waveguide or cavity geometries.

Analysis of Psychological Distress During COVID-19 Among Professionals

The aftermath of the lockdown caused by the current pandemic generates many challenges and opportunities for the professionals as well as for organizations. Several organizations forced the people to work on-site whereas many of the organizations have been allowing work from home. However, both ways of working are challenging and cause psychological distress. The present work analyses the psychological distress among professionals residing in India during the COVID-19 pandemic. The work considers both the scenarios of working professionals: professionals working from home and professionals working onsite. The work introduces a novel hybrid machine learning approach called GBETRR. GBETRR combines two approaches, namely gradient-boosting classifier and extra-trees regressor repressor. The present work also uses a hybrid parameter optimization algorithm. Multiple performance metrics are used to evaluate the performance evaluation. Results revealed that the professionals with work from home are more stressed as compared to the professionals working onsite.

Plasticity of salmonfly (Pteronarcys californica) respiratory phenotypes in response to changes in temperature and oxygen.

Like all taxa, populations of aquatic insects may respond to climate change by evolving new physiologies or behaviors, shifting their range, exhibiting physiological and behavioral plasticity, or going extinct. We evaluated the importance of plasticity by measuring changes in growth, survival and respiratory phenotypes of salmonfly nymphs (the stonefly Pteronarcys californica) in response to experimental combinations of dissolved oxygen and temperature. Overall, smaller individuals grew more rapidly during the 6-week experimental period, and oxygen and temperature interacted to affect growth in complex ways. Survival was lower for the warm treatment, although only four mortalities occurred (91.6% versus 100%). Nymphs acclimated to warmer temperatures did not have higher critical thermal maxima (CTmax), but those acclimated to hypoxia had CTmax values (in normoxia) that were higher by approximately 1°C. These results suggest possible adaptive plasticity of systems for taking up or delivering oxygen. We examined these possibilities by measuring the oxygen sensitivity of metabolic rates and the morphologies of tracheal gill tufts located ventrally on thoracic segments. Mass-specific metabolic rates of individuals acclimated to warmer temperatures were higher in acute hypoxia but lower in normoxia, regardless of their recent history of oxygen exposure during acclimation. The morphology of gill filaments, however, changed in ways that appeared to depress rates of oxygen delivery in functional hypoxia. Our combined results from multiple performance metrics indicate that rising temperatures and hypoxia may interact to magnify the risks to aquatic insects, but that physiological plasticity in respiratory phenotypes may offset some of these risks.

Intrusion Detection Framework for the Internet of Things Using a Dense Random Neural Network

The Internet of Things (IoT) devices, networks, and applications have become an integral part of modern societies. Despite their social, economic, and industrial benefits, these devices and networks are frequently targeted by cybercriminals. Hence, IoT applications and networks demand lightweight, fast, and flexible security solutions to overcome these challenges. In this regard, artificial-intelligence-based solutions with Big Data analytics can produce promising results in the field of cybersecurity. This article proposes a lightweight dense random neural network (DnRaNN) for intrusion detection in the IoT. The proposed scheme is well suited for implementation in resource-constrained IoT networks due to its inherent improved generalization capabilities and distributed nature. The suggested model was evaluated by conducting extensive experiments on a new generation IoT security dataset ToN_IoT. All the experiments were conducted under different hyperparameters and the efficiency of the proposed DnRaNN was evaluated through multiple performance metrics. The findings of the proposed study provide recommendations and insights in binary class and multiclass scenarios. The proposed DnRaNN model attained attack detection accuracy of 99.14% and 99.05% for binary class and multiclass classifications, respectively.

Data Acquisition through Mobile Sink for WSNs with Obstacles Using Support Vector Machine

Mobile sink-based data collection in wireless sensor networks has become an attractive research area to mitigate hotspot issues. It further increases the efficiency of the WSN, such as throughput, lifetime, and energy efficiency, while decreasing delay and packet losses. Mobile sink algorithms developed by many researchers in recent years have only contributed to obtain efficient path planning, and only a few researchers have focused on solving the problem of network environment with obstacles. Here, constructing an obstacle-aware path for the mobile sink to collect data in WSN is a challenging issue. In this context, we present the data acquisition through mobile sink for WSNs with obstacles using support vector machine (DAOSVM). The DAOSVM algorithm works in two phases: visiting point selection and path construction. The visiting point selection uses spanning tree approach, and the path selection uses SVM. The computational complexity of the proposed DAOSVM is estimated and compared using the existing techniques, and it is lower. The DAOSVM also outperforms traditional methods concerning multiple performance metrics under various scenarios.

Time-Adaptive Unsupervised Auditory Attention Decoding Using EEG-Based Stimulus Reconstruction.

The goal of auditory attention decoding (AAD) is to determine to which speaker out of multiple competing speakers a listener is attending based on the brain signals recorded via, e.g., electroencephalography (EEG). AAD algorithms are a fundamental building block of so-called neuro-steered hearing devices that would allow identifying the speaker that should be amplified based on the brain activity. A common approach is to train a subject-specific stimulus decoder that reconstructs the amplitude envelope of the attended speech signal. However, training this decoder requires a dedicated 'ground-truth' EEG recording of the subject under test, during which the attended speaker is known. Furthermore, this decoder remains fixed during operation and can thus not adapt to changing conditions and situations. Therefore, we propose an online time-adaptive unsupervised stimulus reconstruction method that continuously and automatically adapts over time when new EEG and audio data are streaming in. The adaptive decoder does not require ground-truth attention labels obtained from a training session with the end-user and instead can be initialized with a generic subject-independent decoder or even completely random values. We propose two different implementations: a sliding window and recursive implementation, which we extensively validate on three independent datasets based on multiple performance metrics. We show that the proposed time-adaptive unsupervised decoder outperforms a time-invariant supervised decoder, representing an important step toward practically applicable AAD algorithms for neuro-steered hearing devices.

SurvBenchmark: comprehensive benchmarking study of survival analysis methods using both omics data and clinical data.

Survival analysis is a branch of statistics that deals with both the tracking of time and the survival status simultaneously as the dependent response. Current comparisons of survival model performance mostly center on clinical data with classic statistical survival models, with prediction accuracy often serving as the sole metric of model performance. Moreover, survival analysis approaches for censored omics data have not been thoroughly investigated. The common approach is to binarize the survival time and perform a classification analysis.Here, we develop a benchmarking design, SurvBenchmark, that evaluates a diverse collection of survival models for both clinical and omics data sets. SurvBenchmark not only focuses on classical approaches such as the Cox model but also evaluates state-of-the-art machine learning survival models. All approaches were assessed using multiple performance metrics; these include model predictability, stability, flexibility, and computational issues. Our systematic comparison design with 320 comparisons (20 methods over 16 data sets) shows that the performances of survival models vary in practice over real-world data sets and over the choice of the evaluation metric. In particular, we highlight that using multiple performance metrics is critical in providing a balanced assessment of various models. The results in our study will provide practical guidelines for translational scientists and clinicians, as well as define possible areas of investigation in both survival technique and benchmarking strategies.

An efficient multi-objective optimization framework to maintain road networks based on Bayesian optimization and reliability theory

This study proposes an efficient asset management framework that enables decision makers to prioritize the maintenance of their roads, while focusing on the most critical road segments. In particular, this study first extends the application of reliability theory to estimate the overall network condition. Following that, this study proposes a new consequence of failure function for the whole road network based on road segments’ reliability, age, and road agency preferences. Finally, the study proposes an efficient multi-objective optimization algorithm, with the goal of maximizing overall network performance with the least maintenance and computational cost. The suggested framework was applied to three main roads in Jordan and validated statistically by comparing its performance to that of a typical multi-objective genetic algorithm (GA) under various scenarios and utilizing multiple performance metrics. The results show that the proposed algorithm outperforms the traditional GA in terms of objective function values, convergence, and solution spread.

On the value of a multistage optimization approach for intensity-modulated radiation therapy planning* *This work has been conducted under a Lyle School (SMU) Interdisciplinary Research Seed Funding Grant and Radiation Oncology MAIA Lab (UTSW) Educational Experience Agreement (UTSW Contract ID # 2020–7031).

Objective. Intensity-modulated radiation therapy (IMRT) aims to distribute a prescribed dose of radiation to cancerous tumors while sparing the surrounding healthy tissue. A typical approach to IMRT planning uniformly divides and allocates the same dose prescription (DP) across several successive treatment sessions. A more flexible fractionation scheme would lend the capability to vary DPs and utilize updated CT scans and future predictions to adjust treatment delivery. Therefore, our objective is to develop optimization-based models and methodologies that take advantage of adapting treatment decisions across fractions by utilizing predictions of tumor evolution. Approach. We introduce a nonuniform generalization of the uniform allocation scheme that does not automatically assume equal DPs for all sessions. We develop new deterministic and stochastic multistage optimization-based models for such a generalization. Our models allow us to simultaneously identify optimal DPs and fluence maps for individual sessions. We conduct extensive numerical experiments to compare these models using multiple metrics and dose-volume histograms. Main results. Our numerical results in both deterministic and stochastic settings reveal the restrictive nature of the uniform allocation scheme. The results also demonstrate the value of nonuniform multistage models across multiple performance metrics. The improvements can be maintained even when restricting the underlying fractionation scheme to small degrees of nonuniformity. Significance. Our models and computational results support multistage stochastic programming (SP) methodology to derive ideal allocation schemes and fluence maps simultaneously. With technological and computational advancements, we expect the multistage SP methodologies to continue to serve as innovative optimization tools for radiation therapy planning applications.