Reliable Decision Research Articles

A value-based deep reinforcement learning model with human expertise in optimal treatment of sepsis

Deep Reinforcement Learning (DRL) has been increasingly attempted in assisting clinicians for real-time treatment of sepsis. While a value function quantifies the performance of policies in such decision-making processes, most value-based DRL algorithms cannot evaluate the target value function precisely and are not as safe as clinical experts. In this study, we propose a Weighted Dueling Double Deep Q-Network with embedded human Expertise (WD3QNE). A target Q value function with adaptive dynamic weight is designed to improve the estimate accuracy and human expertise in decision-making is leveraged. In addition, the random forest algorithm is employed for feature selection to improve model interpretability. We test our algorithm against state-of-the-art value function methods in terms of expected return, survival rate, action distribution and external validation. The results demonstrate that WD3QNE obtains the highest survival rate of 97.81% in MIMIC-III dataset. Our proposed method is capable of providing reliable treatment decisions with embedded clinician expertise.

Multi-Vehicle Conflict Management With Status and Intent Sharing Under Time Delays

This paper discusses conflict analysis for multiple vehicles possessing different automation levels. Two different classes of cooperation, enabled by vehicle-to-everything (V2X) communication, are considered: status-sharing and intent-sharing. Status sharing allows vehicles to exchange their instantaneous states with each other (e.g., current velocity and position), whereas intent sharing also enables information exchange regarding the future motion of vehicles (e.g., velocity and acceleration bounds). We consider two types of time delays, one in vehicle dynamics and the other in V2X communication. Using reachability theory, our conflict analysis framework interprets the information encoded in the wireless messages pertaining to the two different cooperation classes by means of conflict charts. These charts allow for efficient and reliable on-board decision making and control design for a connected automated ego vehicle which interacts with multiple connected remote vehicles when carrying out different maneuvers. We quantify the effects of time delays in a mixed-autonomy traffic environment and unveil the benefits of intent information. Utilizing real highway data, we use numerical simulations to validate the extended conflict management framework.

Uncertainty-Aware Multiview Deep Learning for Internet of Things Applications

As an essential approach in many Internet of Things (IoT) applications, multiview learning synthesizes multiple features to achieve more comprehensive descriptions of data items. Most of the previous studies on multiview learning have been dedicated to increasing the prediction accuracy, while ignoring the reliability of the decision. This would limit their deployment in high-risk IoT and industrial applications such as the automated vehicle. Although a trusted multiview classification model has been proposed recently, it cannot well deal with the highly complementary multiview data. In this work, we present an evidential multiview deep learning (EMDL) method to make reliable decisions. EMDL first seeks view-specific evidence of each category, which could be termed as the amount of support to each category collected from data. It then dynamically fuses different views at the evidence level to construct the multiview common evidence and makes reliable prediction accordingly (strong evidence indicates high prediction confidence). In particular, we establish a degradation layer to learn the mappings from the common evidence (comprehensive information) to view-specific evidences (partial information) for evidence fusion. It aims to explicitly model consistent and complementary relations in multiview data at the evidence level. We apply EMDL on a synthetic toy dataset and five real-world datasets (three datasets are related to industrial scenarios). Experiments show that EMDL outperforms state-of-the-art baseline methods.

Comparative Analysis of Deep Learning Convolutional Neural Networks based on Transfer Learning for Pneumonia Detection

Abstract: Artificial intelligence has been used in many different fields throughout its development, especially in recent years as the amount of data available has increased. Its major objective is to aid individuals in making more reliable decisions more quickly. Machine learning and artificial intelligence are being used more and more in medicine. This is particularly true in the medical field where a large number of digital files must be collected and processed to apply a variety of biomedical imaging and diagnostic techniques. Machine learning is used to analyse medical images, which helps with consistency and increases reporting accuracy. To process chest X-ray data and assist in diagnosis, this study compares transfer learning-based deep neural networks. The study focuses on comparing and evaluating deep learning methods based on convolutional neural networks for the identification of pneumonia. To establish the model that would accurately diagnose pneumonia, the researchers developed several different models. Five distinct CNN models, including VGG19, VGG16, ResNet50, InceptionNet v3, and YOLO v5, were trained using the RSNA Pneumonia Detection Challenge dataset. Validation Accuracy and Area Under Curve were used to gauge their performance. On test data, VGG16 had the highest validation accuracy (88%) and AUC-ROC (91.8%), whereas YOLO v5 was used to locate the inflammation with a 99% level of confidence.

The Effect of Diurnal Variation on Laboratory Tests

Objective: Commonly used biochemical tests in blood samples may be measured at any time of day. This study investigated the existence and clinical significance of diurnal variations in some of routine parameters to facilitate accurate and reliable decision-making in diagnosis and follow-up.
 Materials and Methods: Blood samples were collected from 17 healthy volunteers who were 18-50 years of age (11 men, 6 women) on the same day at 9:00 am, 12:00 am, 3:00 pm, 6:00 pm, and 12:00 pm. Samples collected at 9:00 am were regarded as baseline. The results of 19 biochemical parameters in blood samples obtained at 12.00 am, 3.00 pm, 6.00 pm and 12.00 pm were statistically and clinically compared with the results at 9.00 am baseline sample. 
 Results: Total protein, creatinine, aspartate transaminase, alanine transaminase, alkaline phosphatase and gamma glutamyl transferase showed no clinically significant variation within the day, but clinically significant changes were observed in levels of glucose, total cholesterol, HDL-cholesterol, triglyceride, total bilirubin (TBIL), direct bilirubin (DBIL), albumin, blood urea nitrogen, uric acid, sodium, potassium, chloride and amylase. Especially, BUN changed by maximum 20-30%, TBIL, DBIL and triglyceride maximum 40-50% within the day.
 Conclusion: The results of our study suggest that clinicians should consider the timing of blood sampling and the diurnal variations in BUN, TBIL, DBIL and triglyceride parameters during diagnosis and treatment follow-up. Sampling throughout the day seems to pose no problem for other tests with limited diurnal variation.

The Application of Artificial Intelligence in Health Care Resource Allocation Before and During the COVID-19 Pandemic: Scoping Review

Background Imbalanced health care resource distribution has been central to unequal health outcomes and political tension around the world. Artificial intelligence (AI) has emerged as a promising tool for facilitating resource distribution, especially during emergencies. However, no comprehensive review exists on the use and ethics of AI in health care resource distribution. Objective This study aims to conduct a scoping review of the application of AI in health care resource distribution, and explore the ethical and political issues in such situations. Methods A scoping review was conducted following the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews). A comprehensive search of relevant literature was conducted in MEDLINE (Ovid), PubMed, Web of Science, and Embase from inception to February 2022. The review included qualitative and quantitative studies investigating the application of AI in health care resource allocation. Results The review involved 22 articles, including 9 on model development and 13 on theoretical discussions, qualitative studies, or review studies. Of the 9 on model development and validation, 5 were conducted in emerging economies, 3 in developed countries, and 1 in a global context. In terms of content, 4 focused on resource distribution at the health system level and 5 focused on resource allocation at the hospital level. Of the 13 qualitative studies, 8 were discussions on the COVID-19 pandemic and the rest were on hospital resources, outbreaks, screening, human resources, and digitalization. Conclusions This scoping review synthesized evidence on AI in health resource distribution, focusing on the COVID-19 pandemic. The results suggest that the application of AI has the potential to improve efficacy in resource distribution, especially during emergencies. Efficient data sharing and collecting structures are needed to make reliable and evidence-based decisions. Health inequality, distributive justice, and transparency must be considered when deploying AI models in real-world situations.

High-frequency forecasting of the crude oil futures price with multiple timeframe predictions fusion

In the abundant literature about crude oil futures price forecasting, researchers generally predicted the crude oil price movements from the perspective of only a single timeframe. In addition, the trading strategies of their trading models were generally designed to be less sophisticated, and their prediction models lacked interpretability. To fill these gaps, a price direction fused prediction and trading approach has been proposed for high-frequency prediction of the Chinese crude oil futures. In the proposed approach, the MTXGBoost (Multiple Timeframes eXtreme Gradient Boosting) is developed and utilized for predictions fusion under multiple timeframes, and the NSGA-II (Non-dominated Sorting Genetic Algorithm-II) is integrated for trading strategy optimization. Moreover, the SHAP (Shapley Additive exPlanation) approach is also employed to interpret how the proposed approach made predictions. Experimental results show that the approach proposed in this research averagely produced a direction prediction accuracy of 78.69%, an accumulated return of 23.17%, and a maximum drawdown of 1.00%, demonstrating that it can produce an excellent profit with small trading risks. Therefore, the proposed approach can be employed as an intelligent, efficient, and reliable decision support system for market investors, energy-related companies, and government departments to make crude oil related decisions.

A Three-Step Framework for Multimodal Industrial Process Monitoring Based on DLAN, TSQTA, and FSBN

The process monitoring method for industrial production can technically achieve early warning of abnormal situations and help operators make timely and reliable response decisions. Because practical industrial processes have multimodal operating conditions, the data distributions of process variables are different. The different data distributions may cause the fault detection model to be invalid. In addition, the fault diagnosis model cannot find the correct root cause variable of system failure by only identifying abnormal variables. There are correlations between the trend states of the process variables. If we do not consider these correlations, this may result in an incorrect fault root cause. Therefore, multimodal industrial process monitoring is a tough issue. In this paper, we propose a three-step framework for multimodal industrial process monitoring. The framework aims for multimodal industrial processes to detect the faulty status timely and then find the correct root variable that causes the failure. We present deep local adaptive network (DLAN), two-stage qualitative trend analysis (TSQTA), and five-state Bayesian network (FSBN) to implement fault detection, identification, and diagnosis step by step. This framework can detect the system failure timely, identify abnormal variables, and find the root cause variable and the fault propagation path. The case studies on the Tennessee Eastman simulation and a practical chlorobenzene production process are provided to verify the effectiveness of the proposed framework in multimodal industrial process monitoring.

Performance Evaluation of Different Decision Fusion Approaches for Image Classification

Image classification is one of the major data mining tasks in smart city applications. However, deploying classification models that have good generalization accuracy is highly crucial for reliable decision-making in such applications. One of the ways to achieve good generalization accuracy is through the use of multiple classifiers and the fusion of their decisions. This approach is known as “decision fusion”. The requirement for achieving good results with decision fusion is that there should be dissimilarity between the outputs of the classifiers. This paper proposes and evaluates two ways of attaining the aforementioned dissimilarity. One is using dissimilar classifiers with different architectures, and the other is using similar classifiers with similar architectures but trained with different batch sizes. The paper also compares a number of decision fusion strategies.

A scoring system for cochlear implant candidate selection using artificial intelligence

Objective: Cochlear implant (CI) candidate selection is a lengthy, complicated process that entails subjective judgment on the interaction of multiple pre-operative variables. It is assumed that setting a scoring system for the process of CI candidate selection would help in precise and reliable decision making. This would also provide a tool that would help in providing a better quality of life for CI patients. Methods: Retrospective cohort study was held out in three post-CI rehabilitation centers. A total of 100 children records were analyzed with two statistical methods; conventional and Artificial Intelligence (AI) using Machine Learning. Language age deficit, phonological deficit, and social deficit were invented as new measures of CI performance; used to represent the developmental delay of those children in a single numeric value (in months). Results: Artificial Intelligence analysis surpassed conventional statistical methods for the prediction of the outcome measures of post-CI performance. This was clearly expressed using linear regression models. The AI classification model validation for predictive accuracy of language age deficit, phonological deficit, and social deficit were 56.66%, 88.11%, and 40.46% respectively. Conclusion: The production of a preliminary CI scoring model used for prediction of performance of patients was achieved. More data should be collected and fed to the software in order to improve its performance.

An improved multiplicative acceptability consistency-driven group decision making with triangular fuzzy reciprocal preference relations

Triangular fuzzy reciprocal preference relation (TFRPR) is one of the most effective tools to express the vague and uncertain preferences of decision makers in group decision making (GDM). Consistency of TFRPRs is a crucial premise for reasonable and reliable decision-making. To do so, this study develops an improved multiplicative acceptability consistency-driven GDM with TFRPRs. By analyzing the composition of TFRPRs, a multiplicative consistency index and the corresponding threshold are defined to measure whether a TFRPR is acceptably consistent. The consistency measurement reflects the essential traits of triangular fuzzy numbers, which fully considers the multiplicative consistency of modal values and the multiplicative consistency of geometric mean based on central tendency. Subsequently, three inconsistent TFRPR cases are analyzed, and the algorithm based on mathematical derivation and a linear programming model is presented to repair the inconsistency of TFRPRs. As per the multiplicative acceptability consistency degree (MACD) of TFRPRs, an MACD-induced ordered weighted averaging (MACD-IOWA) operator is proposed to aggregate individual TFRPRs into group TFRPR. The essence principle is that the larger the MACD value, the higher the decision maker’s weight. Finally, a case study, comparative analysis and discussions are given to verify the feasibility and validity of the proposed methods.

A consolidated decision-making framework for nano-additives selection in battery thermal management applications

The selection of appropriate nano-additives for optimum battery thermal management system (BTMS) performance is an exigent task for decision-makers and is a multi-attribute decision-making (MADM) problem. This work presents a framework that utilizes the application of various MADM schemes. These techniques are integrated with a subjective weighting criteria approach to prioritize suitable nano-additives as promoters towards heat transfer enhancement to mitigate the thermal issues in energy storage devices. The COPRAS, PROMETHEE-II, and VIKOR methods are used to present ranking schemes for alternatives based on the criteria weights assigned by the AHP method. The results show that Graphene and MWCNT nano-additives are the best-suited alternatives. The sensitivity tests are carried out to describe the effect of parametric variations in the most critical risk preferences over the rankings of the alternatives. In addition, the problem of rank reversal associated with these methods is also addressed. Therefore, the evaluation criteria and methodology in this work can provide a theoretical reference for nano-additive selection. An attempt is made to discuss the technical index of efficiency assisting decision-makers to make reliable and robust decisions by employing the CCR model.

Hierarchical policy for seismic intervention of school buildings at urban scale

Seismic risk assessment involves many uncertainties starting from the exposure model and the evaluation of site-specific seismic hazard, and ending with the actual estimation of the expected damages and losses. Seismic risk is also explicitly related to a reliable decision-making process in order to improve pre- and post-earthquake management and resilience. In the case of seismic risk assessment of a large number of critical buildings, like school buildings, at community or regional scale, it is important to establish a reliable and efficient hierarchical policy in order to allow the decision-making regarding the necessary pre- and post- seismic event retrofitting and structural strengthening actions to be undertaken, starting from the identification of the most vulnerable buildings, which need a second order seismic analysis prior to any retrofitting and strengthening action. To this regard, the aim of the present paper is to develop an adequate retrofit optimization framework based on a two-level ranking process for the quick identification of the most vulnerable school buildings, for which a more detailed vulnerability and risk assessment should follow. The herein proposed framework is an efficient and simple tool for prioritization, policy-making and scheduling of seismic prevention projects. The proposed methodology is applied to the school buildings of the municipality of Thessaloniki, Greece. The results of the application indicate that about 3.5% of the total studied school buildings may require further investigation for retrofitting and strengthening.

Probabilistic approach of pre-estimating life-cycle costs of road tunnels

Conceptual pre-estimation of road tunnel costs is a vital yet challenging process during feasibility studies due to the prevailing underground uncertainties and risks. Most existing cost estimation approaches are deterministic, probabilistic methods are limited to a single cost examination. This paper introduces a probabilistic model for pre-estimating road tunnels life-cycle costs. It aims to capture their inherent uncertainty holistically, thus enabling more reliable decision-making at a project’s early stages. The proposed model is developed in three steps: multiple regression analysis, fitting distribution, and Monte-Carlo simulation. The first step unveils the correlations between independent and dependent variables, while the other two steps return the probabilistic descriptions of cost drivers and life-cycle costs. Civil engineering, electromechanical works costs, energy consumption, and operation & maintenance costs are determined as the model’s output variables. A real-world database of 32 dual-bore road tunnels with a total length of 55 km is used to develop probability distributions within reasonable confidence intervals and a sensitivity analysis was performed utilizing tornado charts. The results can assist public authorities and managers in estimating the probability of a specific life-cycle cost and making appropriate decisions in accordance with their risk tolerance.

Towards a Research Agenda for Increasing Trust in Maps and Their Trustworthiness

In this paper a proposal is presented on how cartographers can ensure and/or enhance the trustworthiness of maps. Maps, like all other means of communication, are a result of many decisions. Many of those decisions allow for several options. Default, clear or good practices, guidelines or recommendations do not always exist for which option to choose. It is rather a characteristic of cartography that individual decisions can be made, especially on data preparation, selection, and design aspects.Because every map is a result of many decisions, trust in maps depends on those decisions being reasonable, not questioned, transparently accessible or underpinned by reliable decision makers (reliable in terms of their record as an institution, because of their reputation, because of their competence).The advent of easy-to-use map making software and data handling instruments put some of those decisions in the hands of "everyone", as well as in the hands of many who might have an interest in using maps as instruments of communicating a specific agenda.This is like other means of communication and has led to a rising discussion on "fake news", "fake media" and "fake maps", thus ultimately how we can trust the communicated information and how we can distinguish "fake" from "trustworthy" maps.Therefore, in this paper a proposal is presented on how cartographers can ensure and/or enhance the trustworthiness of maps. This is done by introducing two concepts of "going deep", allowing for transparency of cartographic decisions and "going wide", allowing for contextualizing a given map by being able to access alternative maps of the same topic.This proposal adds to the existing attempts to contribute to an ethical framework for map making, such as the Locus Charter or the Code of Ethics of the British Cartographic Society.

Trustworthy Federated Learning via Blockchain

The safety-critical scenarios of artificial intelligence (AI), such as autonomous driving, Internet of Things, smart healthcare, etc., have raised critical requirements of trustworthy AI to guarantee the privacy and security with reliable decisions. As a nascent branch for trustworthy AI, federated learning (FL) has been regarded as a promising privacy preserving framework for training a global AI model over collaborative devices. However, security challenges still exist in the FL framework, e.g., Byzantine attacks from malicious devices, and model tampering attacks from malicious server, which will degrade or destroy the accuracy of trained global AI model. In this article, we shall propose a decentralized blockchain-based FL (B-FL) architecture by using a secure global aggregation algorithm to resist malicious devices, and deploying a practical Byzantine fault tolerance consensus protocol with high effectiveness and low energy consumption among multiple edge servers to prevent model tampering from the malicious server. However, to implement B-FL system at the network edge, multiple rounds of cross-validation in blockchain consensus protocol will induce long training latency. We thus formulate a network optimization problem that jointly considers bandwidth and power allocation for the minimization of long-term average training latency consisting of progressive learning rounds. We further propose to transform the network optimization problem as a Markov decision process and leverage the deep reinforcement learning (DRL)-based algorithm to provide high system performance with low computational complexity. Simulation results demonstrate that B-FL can resist malicious attacks from edge devices and servers, and the training latency of B-FL can be significantly reduced by the DRL-based algorithm compared with the baseline algorithms.

Reliability engineering opportunities in Industry 4.0

Industry 4.0 is based on the internet of things, this means, that in industries or companies the value chain, including processes, products, hardware, and software, among others, exists the necessity to implement new knowledge technologies to implement and control Industry 4.0 interactions, in other words, with the advantages of industry 4.0 such as the used of real-time data, big data, blockchain, human-machine interaction, cyber-systems, among others, the electronic devices that allowed this interaction and electronic data interchange centralize, need to be in the highest possible reliability perform, aiming to be able to carry out all these activities safely and effectively. In this research, the objective is to introduce some general advantages of Industry 4.0 and how reliability engineering is an important ally in the performance of the functions of Industry 4.0. Then, this manuscript presents the necessary knowledge and applications of reliability engineering that can be implemented in Industry 4.0 as stress-strength analysis in the case of equal shape parameters βs =βS; stress-strength analysis βs ≠ βS; Nonnormal capability index and Weibull Capability index, by using Weibull++ Software and mathematical procedure. Another target of this research is to demonstrate that by using reliability engineering it is possible to have better control in such a way that efficiency and productivity stay and even can increases. Also, reliability engineering represents security and stable processes. Finally, with the merging of Industry 4.0 and reliability engineering, the making decision process is more reliable decision-making.

Robust learning and optimization in distributionally robust stochastic variational inequalities under uncertainty

Robust learning and optimization in distributionally robust stochastic variational inequalities under uncertainty is a crucial research area that addresses the challenge of making optimal decisions in the presence of distributional ambiguity. This research explores the development of methodologies and algorithms to handle uncertainty in variational inequalities, incorporating a distributionally robust framework that considers a range of possible distributions or uncertainty sets. By minimizing the worst-case expected performance across these distributions, the proposed approaches ensure robustness and optimality in decision-making under uncertainty. The research encompasses theoretical analysis, algorithm development, and empirical evaluations to demonstrate the effectiveness of the proposed methodologies in various domains, such as portfolio optimization and supply chain management. The outcomes of this research contribute to the advancement of robust optimization techniques, enabling decision-makers to make reliable and robust decisions in complex real-world systems

Multi criteria group decision (MCGDM) for selecting third-party logistics provider (3PL) under Pythagorean fuzzy rough Einstein aggregators and entropy measures

<abstract> <p>In real life, with the trend of outsourcing logistics activities, choosing a third-party logistics (3PL) provider has become an inevitable choice for shippers. One of the most difficult decisions logistics consumers are facing the selecting the 3PL provider that best meets their needs. Decision making (DM) is an important in dealing with such situations because it allows them to make reliable decisions in a short period of time, as incorrect decisions can result in huge financial losses. In this regard, this article provides a new multi criteria group decision making method (MCGDM) under Pythagorean fuzzy rough (PyFR) set. A series of new PyFR Einstein weighted averaging aggregation operators and their basic aspects are described in depth. To evaluate the weights of decision experts and criteria weights we established the PyFR entropy measure. Further, using multiple aggregation methods based on PyFR information, a novel algorithm is offered to solve issues with ambiguous or insufficient data to obtain reliable and preferable results. First, decision-experts use PyFR sets to represent their evaluation information on alternatives based on the criteria. Then, apply all these proposed PyFR Einstein aggregation lists to rank all alternatives and find the best optimal result. Finally, to demonstrate the feasibility of the proposed PyFR decision system, a real example of choosing a 3PL is given.</p> </abstract>

Use of imaging modalities for decision-making in inflammatory bowel disease.

Cross-sectional magnetic resonance enterography (MRE) and intestinal ultrasonography (IUS) provide valuable and noninvasive information to accurately assess disease activity, severity, and extent; detect complications; and monitor the response to treatment, as well as predict the postoperative recurrence of Crohn's disease and a negative disease course. Therefore, both imaging modalities are emerging as pivotal diagnostic tools to achieve the emerging therapeutic target of transmural healing associated with better disease outcomes. Despite its numerous potential advantages over endoscopy and even MRE and its good availability, IUS is still widely underused to monitor and manage inflammatory bowel disease (IBD) patients and help in making clinical decisions in routine practice. This situation is clearly due to the absence of validated, reliable, and responsive indices, as well as the lack of trained gastroenterologists and radiologists, as IUS is a component of radiologist expertise in several countries but not yet integrated into the training program of gastroenterologists. However, there is an increasing body of evidence in the literature that IUS and MRE are both becoming essential imaging resources to help clinicians in making reliable decisions. Here, we discuss the up-to-date evidence about the usefulness and performance of cross-sectional imaging, focusing on the ability of bowel US and MRE to aid clinical decision-making for the optimal management and monitoring of IBD.