A5030661785- Insurance, Mortality, Demography, Risk Management
- Probability and Risk Models
- Insurance and Financial Risk Management
- Lung Cancer Treatments and Mutations
- Lung Cancer Diagnosis and Treatment
- demographic modeling and climate adaptation
- Stochastic processes and statistical mechanics
- Lung Cancer Research Studies
- Bayesian Methods and Mixture Models
- Markov Chains and Monte Carlo Methods
- Mathematical and Theoretical Epidemiology and Ecology Models
- COVID-19 epidemiological studies
- Global Health Care Issues
- Geriatric Care and Nursing Homes
- Atrial Fibrillation Management and Outcomes
- Financial Distress and Bankruptcy Prediction
- AI in cancer detection
- Esophageal Cancer Research and Treatment
- Complex Network Analysis Techniques
- Healthcare Operations and Scheduling Optimization
- Healthcare innovation and challenges
- Retirement, Disability, and Employment
- Artificial Intelligence in Healthcare and Education
- SARS-CoV-2 and COVID-19 Research
- Educational Technology and Optimization
Universidade Nova de Lisboa
2013-2024
Instituto Superior de Ciências da Informação e da Administração
2023
University of Lisbon
2008-2023
Hospital Universitario Puerta de Hierro Majadahonda
2022
Sociedade Brasileira de Educação Matemática
2013
Artificial intelligence (AI) has contributed substantially in recent years to the resolution of different biomedical problems, including cancer. However, AI tools with significant and widespread impact oncology remain scarce. The goal this study is present an AI-based solution tool for cancer patients data analysis that assists clinicians identifying clinical factors associated poor prognosis, relapse survival, develop a prognostic model stratifies by risk.We used from 5275 diagnosed...
Under the assumptions of an open portfolio, i.e., considering that a policyholder can transfer his policy to another insurance company and continuous arrival new policyholders into portfolio which be placed any bonus classes not only in "starting class", we developed model (Stochastic Vortices Model) estimate Long Run Distribution for Bonus Malus System.These hypothesis render quite representative reality.With obtained Distribution, few optimal scales were calculated, such as Norberg's...
We propose a method for fitting transition intensities to sufficiently large set of trajectories continuous-time non-homogeneous Markov chain with finite state space. Starting simulated data computed Gompertz–Makeham intensities, we apply the proposed fit piecewise linear and then compare probabilities corresponding both fitted intensities; main comparison result is that order magnitude average error per unit time—chosen as year—is always less than 1%, thus validating methodology proposed.
Modelling claim frequency and severity are topics of great interest in property-casualty insurance for supporting underwriting, ratemaking, reserving actuarial decisions. Standard Generalized Linear Models (GLM) frequency–severity models assume a linear relationship between function the response variable predictors, independence severity, assign full credibility to data. To overcome some these restrictions, this paper investigates predictive performance Gradient Boosting with decision trees...
Abstract Motor insurance is a very competitive business where insurers operate with quite large portfolios, often decisions must be taken under short horizons and therefore ruin probabilities should calculated in finite time. The probability of ruin, continuous time, numerically evaluated the classical Cramér–Lundberg risk process framework for motor portfolio, we allow posteriori premium adjustments, according to claim record each individual policyholder. Focusing on model bonus-malus...
Abstract In this paper, following an open portfolio approach, we show how to estimate a Bonus-malus system evolution. Considering model for the number of new annual policies, obtain ML estimators, asymptotic distributions and confidence regions expected policies entering in each year, as well proportion insureds bonus class, by year enrollment. Confidence distribution policyholders result optimal scales. Our treatment is illustrated example with numerical results.
In this paper, we study, by means of randomized sampling, the long-run stability some open Markov population fed with time-dependent Poisson inputs. We show that state probabilities within transient states converge—even when overall expected dimension increases without bound—under general conditions on transition matrix and input intensities.Following convergence results, obtain ML estimators for a particular sequence intensities, where new arrivals is modeled sigmoidal function. These allow...
We consider a non-homogeneous continuous time Markov chain model for Long-Term Care with five states: the autonomous state, three dependent states of light, moderate and severe dependence levels death state. For general approach, we allow non null intensities all returns from higher to lesser dependencies in multi-state model. Using data 2015 Portuguese National Network Continuous database, as main research contribution this paper, propose method calibrate transition one step probabilities...
We study—with existence and unicity results—a variant of the SIR model for an infectious disease incorporating both possibility a death outcome—in short period time—and regime switch that can account mitigation measures used to control spreading infections, such as total lockdown. This is parametrised by three parameters: basic reproduction number, mortality rate infected, duration disease. discuss particular example application Portuguese COVID-19 data in two periods just after start...
Our paper considers open populations with arrivals and departures whose elements are subject to periodic reclassications. These will be divided into a nite number of sub-populations. Assuming that: a) entries, reclassications occur at the beginning time units; b) reallocated equally spaced times; c) numbers new entering units realizations independent Poisson distributed random variables; we use Markov chains obtain limit results for relative sizes sub-populations corresponding states chain....
Abstract For a large motor insurance portfolio, on an open environment, we study the impact of experience rating in finite and continuous time ruin probabilities. We consider model for calculating probabilities applicable to portfolios with Markovian Bonus‐Malus System (BMS), based claim counts, automobile portfolio using classical risk framework model. New challenges are brought when scenario is introduced. When compared BMS approach may change significantly. By Portuguese insurer,...
We address the problem of finding a natural continuous time Markov type process—in open populations—that best captures information provided by an chain in discrete which is usually sole possible observation from data. Given chain, we single out two main approaches: In first one, consider calibration procedure process using transition matrix and show that, when embeddable has optimal solutions. second approach, semi-Markov processes—and schemes—and propose direct extension theory to one known...
Abstract We present a methodology to connect an ordinary differential equation (ODE) model of interacting entities at the individual level, open Markov chain (OMC) population such individuals, via stochastic (SDE) intermediate model. The ODE here presented is formulated as dynamic change between two regimes; one regime mean reverting type and other inverse logistic type. For general purpose defining OMC for we associate Ito processes, in form SDE system equations, by means addition Gaussian...
Views Icon Article contents Figures & tables Video Audio Supplementary Data Peer Review Share Twitter Facebook Reddit LinkedIn Tools Reprints and Permissions Cite Search Site Citation Manuel L. Esquível, Gracinda R. Guerreiro, José M. Fernandes, Ana F. Silva; On a spread model for portfolio credit risk modeling. AIP Conf. Proc. 10 March 2015; 1648 (1): 540004. https://doi.org/10.1063/1.4912750 Download citation file: Ris (Zotero) Reference Manager EasyBib Bookends Mendeley Papers EndNote...
Current prognosis in oncology is reduced to the tumour stage and performance status, leaving out many other factors that may impact patient´s management. Prognostic stratification of early non-small-cell lung cancer (NSCLC) patients with poor after surgery considerable clinical relevance. The objective this study was identify associated long-term overall survival a real-life cohort I-II NSCLC develop prognostic model identifies features stratifies by risk. This including 505 patients,...
We introduce a schematic formalism for the time evolution of random population entering some set classes and such that each member evolves among these according to scheme based on Markov chain model. consider flow incoming members is modeled by series we detail structure elements in classes. present practical application data from credit portfolio Cape Verdian bank; after modeling two different ways - namely as an ARIMA process deterministic sigmoid type trend plus SARMA residues simulate...
Modelling claim frequency and severity are topics of great interest in property-casualty insurance for supporting underwriting, ratemaking, reserving actuarial decisions. This paper investigates the predictive performance Gradient Boosting with Decision Trees as base learners to model motor using a private cross-country large dataset. The algorithm combines many weak tackle conceptual uncertainty empirical research. findings show that is superior standard Generalised Linear Model sense it...