Credit risk is the potential exposure of a creditor to an obligor's failure or refusal to repay the debt in principal or interest. The potential of exposure is measured in terms of probability of default. Many models have been developed to estimate credit risk, with rating agencies dating back to the 19th century. They provide their assessment of probability of default and transition probabilities of various firms in their annual reports. Regulatory capital requirements for credit risk outlined by the Basel Committee on Banking Supervision have made it essential for banks and financial institutions to develop sophisticated models in an attempt to measure credit risk with higher accuracy. The Bayesian framework proposed in this article uses the techniques developed in physical sciences and engineering for dealing with model uncertainty and expert accuracy to obtain improved estimates of credit risk and associated uncertainties. The approach uses estimates from one or more rating agencies and incorporates their historical accuracy (past performance data) in estimating future default risk and transition probabilities. Several examples demonstrate that the proposed methodology can assess default probability with accuracy exceeding the estimations of all the individual models. Moreover, the methodology accounts for potentially significant departures from "nominal predictions" due to "upsetting events" such as the 2008 global banking crisis.
While predicting child maltreatment risk at the household level is useful for allocating limited child welfare resources, significant privacy, data integration, data governance and legal hurdles make such an algorithm economically and politically difficult to put into production. In this project, we take a different approach to child maltreatment risk prediction, developing machine learning models that predict, not for a household but for a small spatial areal unit, such as the block. The only private health data required for this use case are geocoded maltreatment events. We present the results of a machine learning analysis in Richmond Virginia, including exploratory analysis, feature engineering, model development and validation. We then interpret our models in a resource allocation context.
This article aims to consider whether classic criminal offences (such as manslaughter) are adequate to reprove the scientists' behaviour when major calamities are being judged to have caused the death of people and wide destructions. The fundamental problem hinges on the role of risk-assessment and consultancy carried out by the scientists, as well as on the unknown state of major risks. Then, to establish a link of causality between the defendants' behaviour and the death-events affecting the victims, it must be proved that: a) the scientists "psychically" influenced the victims to leave any safety precaution in relation to the risk; b) the deaths of the inhabitants are not to be considered an "extraordinary" circumstance, even by experts. The difficulties faced by the Judge to fulfil these tasks prompt us to wonder whether other types of criminal charges would be more appropriate for sanctioning scientists who are found to be derelict in their duty of risk-assessment to authorities and citizens.
This study evaluates several feature ranking techniques together with some classifiers based on machine learning to identify relevant factors regarding the probability of contracting breast cancer and improve the performance of risk prediction models for breast cancer in a healthy population. The dataset with 919 cases and 946 controls comes from the MCC-Spain study and includes only environmental and genetic features. Breast cancer is a major public health problem. Our aim is to analyze which factors in the cancer risk prediction model are the most important for breast cancer prediction. Likewise, quantifying the stability of feature selection methods becomes essential before trying to gain insight into the data. This paper assesses several feature selection algorithms in terms of performance for a set of predictive models. Furthermore, their robustness is quantified to analyze both the similarity between the feature selection rankings and their own stability. The ranking provided by the SVM-RFE approach leads to the best performance in terms of the area under the ROC curve (AUC) metric. Top-47 ranked features obtained with this approach fed to the Logistic Regression classifier achieve an AUC = 0.616. This means an improvement of 5.8% in comparison with the full feature set. Furthermore, the SVM-RFE ranking technique turned out to be highly stable (as well as Random Forest), whereas relief and the wrapper approaches are quite unstable. This study demonstrates that the stability and performance of the model should be studied together as Random Forest and SVM-RFE turned out to be the most stable algorithms, but in terms of model performance SVM-RFE outperforms Random Forest. ; The study was partially funded by the "Accion Transversal del Cancer", approved on the Spanish Ministry Council on the 11th October 2007, by the Instituto de Salud Carlos III-FEDER (PI08/1770, PI08/0533, PI08/1359, PS09/00773, PS09/01286, PS09/01903, PS09/02078, PS09/01662, PI11/01403, PI11/01889, PI11/00226, PI11/01810, PI11/02213, PI12/00488, PI12/00265, PI12/01270, PI12/00715, PI12/00150), by the Fundación Marqués de Valdecilla (API 10/09), by the ICGC International Cancer Genome Consortium CLL, by the Junta de Castilla y León (LE22A10-2), by the Consejería de Salud of the Junta de Andalucía (PI-0571), by the Conselleria de Sanitat of the Generalitat Valenciana (AP 061/10), by the Recercaixa (2010ACUP 00310), by the Regional Government of the Basque Country by European Commission grants FOOD-CT- 2006-036224- HIWATE, by the Spanish Association Against Cancer (AECC) Scientific Foundation, by the The Catalan Government DURSI grant 2009SGR1489. Samples: Biological samples were stored at the Parc de Salut MAR Biobank (MARBiobanc; Barcelona) which is supported by Instituto de Salud Carlos III FEDER (RD09/0076/00036). Furthermore, at the Public Health Laboratory from Gipuzkoa and the Basque Biobank. Furthermore, sample collection was supported by the Xarxa de Bancs de Tumors de Catalunya sponsored by Pla Director d'Oncologia de Catalunya (XBTC). Biological samples were stored at the "Biobanco La Fe" which is supported by Instituto de Salud Carlos III (RD 09 0076/00021) and FISABIO biobanking, which is supported by Instituto de Salud Carlos III (RD09 0076/00058). ; Sí
This paper uses the text data mining method to separate the intonation in the annual reports of credit risk enterprises and non-credit risk enterprises, quantify it, and study the impact of annual report intonation on the effectiveness of credit risk prediction. In the empirical research, this paper uses the factor analysis method for some traditional financial variables, and uses the extracted components and intonation variables to predict the credit risk through the logistic model. The results show that the tone of enterprises with credit risk is more negative, and the degree of pessimism is significantly positively correlated with the probability of credit risk. By comparing the ROC curves of the prediction results before and after the addition of intonation variables, adding intonation variables to the credit risk prediction based on financial variables can improve the effectiveness of the prediction.
PurposeProject portfolio risk (PPR) management plays an important role in promoting the smooth implementation of a project portfolio (PP). Accurate PPR prediction helps managers cope with risks timely in complicated PP environments. However, studies on accurate PPR impact degree prediction, which consists of both risk occurrence probabilities and risk impact consequences considering project interactions, are limited. This study aims to model PPR prediction and expand PPR prediction tools.Design/methodology/approachIn this study, the authors build a PPR prediction model based on a genetic algorithm and back-propagation neural network (GA-BPNN) integrated with entropy-trapezoidal fuzzy numbers. Then, the authors verify the proposed model with real data and obtain PPR impact degrees.FindingsThe test results indicate that the proposed method achieves an average absolute error of 0.002 and an average prediction accuracy rate of 97.8%. The former is reduced by 0.038, while the latter is improved by 32.1% when compared with the results of the original BPNN model. Finally, the authors conduct an index sensitivity analysis for identifying critical risks to effectively control them.Originality/valueThis study develops a hybrid PPR prediction model that integrates a GA-BPNN with entropy-trapezoidal fuzzy numbers. The authors use this model to predict PPR impact degrees, which consist of both risk occurrence probabilities and risk impact consequences considering project interactions. The results provide insights into PPR management.
In this chapter we review the epidemiology of lung cancer incidence and mortality among never smokers/nonsmokers and describe the never smoker lung cancer risk models used by the Cancer Intervention and Surveillance Network (CISNET) modelers. Our review focuses on those influences likely to have measurable population impact on never smoker risk, such as secondhand smoke, even though the individual‐level impact may be small. Occupational exposures may also contribute importantly to the population attributable risk of lung cancer. We examine the following risk factors in this chapter: age, environmental tobacco smoke, cooking fumes, ionizing radiation including radon gas, inherited genetic susceptibility, selected occupational exposures, preexisting lung disease, and oncogenic viruses. We also compare the prevalence of never smokers between the three CISNET smoking scenarios and present the corresponding lung cancer mortality estimates among never smokers as predicted by a typical CISNET model.
Rapid diagnostic tools for children with Ebola virus disease (EVD) are needed to expedite isolation and treatment. To evaluate a predictive diagnostic tool, we examined retrospective data (2014-2015) from the International Medical Corps Ebola Treatment Centers in West Africa. We incorporated statistically derived candidate predictors into a 7-point Pediatric Ebola Risk Score. Evidence of bleeding or having known or no known Ebola contacts was positively associated with an EVD diagnosis, whereas abdominal pain was negatively associated. Model discrimination using area under the curve (AUC) was 0.87, which outperforms the World Health Organization criteria (AUC 0.56). External validation, performed by using data from International Medical Corps Ebola Treatment Centers in the Democratic Republic of the Congo during 2018-2019, showed an AUC of 0.70. External validation showed that discrimination achieved by using World Health Organization criteria was similar; however, the Pediatric Ebola Risk Score is simpler to use.
In: Bulletin of the World Health Organization: the international journal of public health = Bulletin de l'Organisation Mondiale de la Santé, Band 101, Heft 4, S. 238-247
In light of the COVID-19 pandemic, the Medicines and Healthcare products Regulatory Agency administered the standards for producing a Rapidly Manufactured Ventilator System (RMVS) free of charge due to the United Kingdom's shortfall of ventilator systems throughout health centers. The standards delineate the minimum requirements in which a Rapidly Manufactured Ventilator System must encompass to be admissible for usage within hospitals. This work commences by evaluating the standards provided by the government to identify any potential security vulnerabilities that may arise due to the succinct development standards provided by the MHRA. This research investigates what cyber considerations are taken to safeguard a patient's health and medical data to improve situational awareness. A tool for a remotely accessible, low-cost ventilator system is developed to reveal what a malicious actor may be able to inflict on a modern ventilator and its adverse impact.
AbstractPredicting terrorism risk is crucial for formulating detailed counter‐strategies. However, this task is challenging mainly because the risk of the concerned potential victim is not isolated. Terrorism risk has a spatiotemporal interprovincial contagious characteristic. The risk diffusion mechanism comes from three possibilities: cross‐provincial terrorist attacks, internal and external echoes, and internal self‐excitation. This study proposed a novel spatiotemporal graph convolutional network (STGCN)‐based extension method to capture the complex and multidimensional non‐Euclidean relationships between different provinces and forecast the daily risks. Specifically, three graph structures were constructed to represent the contagious process between provinces: the distance graph, the province‐level root cause similarity graph, and the self‐excited graph. The long short‐term memory and self‐attention layers were extended to STGCN for capturing context‐dependent temporal characters. At the same time, the one‐dimensional convolutional neural network kernel with the gated linear unit inside the classical STGCN can model single‐node‐dependent temporal features, and the spectral graph convolution modules can capture spatial features. The experimental results on Afghanistan terrorist attack data from 2005 to 2020 demonstrate the effectiveness of the proposed extended STGCN method compared to other machine learning prediction models. Furthermore, the results illustrate the crucial of capturing comprehensive spatiotemporal correlation characters among provinces. Based on this, this article provides counter‐terrorism management insights on addressing the long‐term root causes of terrorism risk and performing short‐term situational prevention.