Showing papers by "Dong-Ling Xu published in 2020"

A Belief Rule-Based Expert System for Fault Diagnosis of Marine Diesel Engines

Abstract: This paper proposes a new belief rule-based (BRB) expert system for fault diagnosis of marine diesel engines. The expert system is the first of its kind that consists of multiple concurrently activated BRB subsystems, in which each subsystem has its distinctive outputs and uses the evidential reasoning approach for inference. This novel modeling approach can be applied to identify fault modes that may co-exist. In essence, the group of BRB subsystems is used to model the nonlinear relationships between the fault features and the fault modes in marine diesel engines. The initial BRB expert system can be established by using expert experience and then optimized by using the data samples accumulated during the operation of marine diesel engines. Due to limitations in knowledge and data collected, ignorance is also considered in some BRB subsystems. The proposed BRB expert system is applied to abnormal wear detection for a kind of marine diesel engine. The performance of the BRB expert system is investigated in comparison with that of artificial neural network (ANN) models, support vector machine (SVM) models, and binary logistic regression model with fivefold cross-validation. The results show that the BRB expert system can be used for fault diagnosis of marine diesel engines in a probabilistic manner, which outperforms the ANN models, SVM models, and the binary logistic regression model in terms of accuracy and stability, and can effectively identify concurrent faults.

An Explainable AI Decision-Support-System to Automate Loan Underwriting

Abstract: Widespread adoption of automated decision making by artificial intelligence (AI) is witnessed due to specular advances in computation power and improvements in optimization algorithms especially in machine learning (ML). Complex ML models provide good prediction accuracy; however, the opacity of ML models does not provide sufficient assurance for their adoption in the automation of lending decisions. This paper presents an explainable AI decision-support-system to automate the loan underwriting process by belief-rule-base (BRB). This system can accommodate human knowledge and can also learn from historical data by supervised learning. The hierarchical structure of BRB can accommodates factual and heuristic rules. The system can explain the chain of events leading to a decision for a loan application by the importance of an activated rule and the contribution of antecedent attributes in the rule. A business case study on automation of mortgage underwriting is demonstrated to show that the BRB system can provide a good trade-off between accuracy and explainability. The textual explanation produced by the activation of rules could be used as a reason for denial of a loan. The decision-making process for an application can be comprehended by the significance of rules in providing the decision and contribution of its antecedent attributes.

An evidential reasoning approach based on risk attitude and criterion reliability

Abstract: In the evidential reasoning (ER) approach, in addition to the weight of a criterion, reliability is another important concept in connection with the criterion although it attracts little attention. Additionally, a decision maker’s risk attitude is of particular importance in decision-making. To simultaneously consider these two important factors, this paper proposes a new ER approach. In the approach, with the consideration of a decision maker’s risk attitude, the combinational reliability of each criterion for each alternative is constructed from the original reliability of the criterion. By following the regression idea to learn the original reliability of each criterion, a unified optimization model is constructed, in which the maximum difference between combinational reliabilities and their estimations is minimized. Within the post-optimal solution space of criterion reliabilities found by solving the unified model, another optimization model is constructed, from which the minimum and maximum expected utilities of each alternative are determined. Solutions to multi-criteria decision-making problems are then generated from the expected utilities. The proposed approach is used to analyze a material supplier selection problem for a company located in Tongling, Anhui, China. The analysis of the problem demonstrates the applicability of the proposed approach.

Evidence updating with static and dynamical performance analyses for industrial alarm system design.

Abstract: In the Dempster-Shafer theory (DST) of evidence, the alarm evidence updating-based method can effectively deal with the uncertainty of the monitored process variable so as to significantly reduce the false alarm rates (FAR) and missed alarm rates (MAR) of the industrial alarm system. But the price of the decrease of FAR and MAR is the increase of the averaged alarm delay (AAD). In order to obtain better comprehensive performance, besides the accuracy indices (FAR and MAR), the sensitivity index (AAD) should be considered simultaneously in the alarm system parameter optimization design. In the framework of DST, firstly, this paper defines the static and dynamical performance indices in the alarm evidence space which are compatible with FAR/MAR/AAD in the process variable space. But the former can measure the performance of the DST-based alarm systems more naturally and elaborately than the latter; secondly, a systematic parameter optimization design procedure for the alarm system is investigated by using these new indices and the tradeoff among them. Finally, two typical numerical experiments and an industrial case are provided to illustrate the effectiveness of the static and dynamical indices for improving the comprehensive performance of the DST-based alarm systems.

Distance-based intuitionistic multiplicative multiple criteria decision-making methods for healthcare management in West China Hospital

Abstract: Funding information National Natural Science Fo undation of China, Grant/Award Numbers: 71532007, 71771156, 71971145; NSFC-Zhejiang Joint Fund for the Integration of Industrialization and Informatization, Grant/Award Number: U1709215 Abstract Intuitionistic multiplicative sets use an asymmetric, unbalanced scale to express information from positive, negative, and indeterminate information. They have been found capable of comprehensively and objectively representing a person's intuitive understanding and hence have attracted much attention. Distance techniques are widely used to measure the degree to which arguments deviate from one another. Several fuzzy set extensions have been developed, but little research has been conducted on measures of distance between intuitionistic multiplicative sets. In this paper, we start by presenting a variety of measures of the distance between intuitionistic multiplicative sets, including Hausdorff distance measures, weighted distance measures, ordered weighted distance measures, and continuous weighted distance measures. We then develop a distance-based intuitionistic multiplicativetechnique for order preference by similarity to ideal solution method and a distancebased intuitionistic multiplicative-Vlsekriterijumska Optimizacija I Kompromisno Resenje method for handling multiple criteria decision-making problems with intuitionistic multiplicative evaluation information. To demonstrate the practical application of these distance measures and the proposed methods, we provide a case study of hospital management of inpatient admission. The paper ends with comparative analyses of the two methods and some concluding remarks.

Automobile Insurance Fraud Detection using the Evidential Reasoning Approach and Data-Driven Inferential Modelling

Abstract: Automobile insurance fraud detection has become critically important for reducing the costs of insurance companies. The majority of insurance companies use expert knowledge to detect fraud. Experience-based knowledge are interpretable and re-usable but the simplistic way that this knowledge is used in practice, often leads to some degree of misjudgment. This paper aims to establish a unique Evidential Reasoning (ER) rule that combines independent evidence from both experience based indicators and probabilities of fraud obtained from historical data. Each piece of evidence is weighted and then combined conjunctively with the weights optimised using a maximum likelihood evidential reasoning (MAKER) framework for datadriven inferential modelling. Based on a real-world insurance claim dataset, our experimental results reveal that the proposed approach preserves the interpretability and usability of expert detection system, and anticipates the changes in fraud practices by tracking the trend of the weights of experience-based indicators. Furthermore, the experimental results show that the proposed approach outperforms a number of widely used machine learning models, such as logistic regression and random forests.