AGGREGATE CLAIM ESTIMATION USING BIVARIATE HIDDEN MARKOV MODEL

In this paper, we propose an approach for modeling claim dependence, with the assumption that the claim numbers and the aggregate claim amounts are mutually and serially dependent through an underlying hidden state and can be characterized by a hidden finite state Markov chain using bivariate Hidden Markov Model (BHMM). We construct three different BHMMs, namely Poisson–Normal HMM, Poisson–Gamma HMM, and Negative Binomial–Gamma HMM, stemming from the most commonly used distributions in insurance studies. Expectation Maximization algorithm is implemented and for the maximization of the state-dependent part of log-likelihood of BHMMs, the estimates are derived analytically. To illustrate the proposed model, motor third-party liability claims in Istanbul, Turkey, are employed in the frame of Poisson–Normal HMM under a different number of states. In addition, we derive the forecast distribution, calculate state predictions, and determine the most likely sequence of states. The results indicate that the dependence under indirect factors can be captured in terms of different states, namely low, medium, and high states.

[1]  D. Haussler,et al.  Hidden Markov models in computational biology. Applications to protein modeling. , 1993, Journal of molecular biology.

[2]  Conditional Least Squares and Copulae in Claims Reserving for a Single Line of Business , 2013, 1306.4529.

[3]  Jr. G. Forney,et al.  The viterbi algorithm , 1973 .

[4]  K. Yuen,et al.  On a correlated aggregate claims model with thinning-dependence structure , 2005 .

[5]  Jean Pinquet,et al.  Experience Rating through Heterogeneous Models , 2000 .

[6]  Jan Dhaene,et al.  On the dependency of risks in the individual life model , 1997 .

[7]  Christian Genest,et al.  Generalized linear models for dependent frequency and severity of insurance claims , 2015 .

[8]  Leo Wang-Kit Cheung,et al.  Use of Runs Statistics for Pattern Recognition in Genomic DNA Sequences , 2004, J. Comput. Biol..

[9]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[10]  Peter Guttorp,et al.  A Hidden Markov Model for Space‐Time Precipitation , 1991 .

[11]  José G. Dias,et al.  Dynamic clustering of energy markets: An extended hidden Markov approach , 2014, Expert Syst. Appl..

[12]  X. Lin,et al.  A Marked Cox Model for the Number of IBNR Claims: Theory , 2016 .

[13]  D. Rubin,et al.  Statistical Analysis with Missing Data. , 1989 .

[14]  Yiu Kuen Tse,et al.  Nonlife Actuarial Models: Theory, Methods and Evaluation , 2009 .

[15]  Biing-Hwang Juang,et al.  Fundamentals of speech recognition , 1993, Prentice Hall signal processing series.

[16]  S. Utev,et al.  Measuring the impact of dependence between claims occurrences , 2002 .

[17]  Jiandong Ren.  A multivariate aggregate loss model , 2012 .

[18]  Lawrence R. Rabiner,et al.  A tutorial on hidden Markov models and selected applications in speech recognition , 1989, Proc. IEEE.

[19]  M. Pevsta,et al.  Modeling Dependencies in Claims Reserving with GEE , 2013, 1306.3768.

[20]  Andrew J. Viterbi,et al.  Error bounds for convolutional codes and an asymptotically optimum decoding algorithm , 1967, IEEE Trans. Inf. Theory.

[21]  Alfred Müller,et al.  MODELING AND COMPARING DEPENDENCIES IN MULTIVARIATE RISK PORTFOLIOS , 1998 .

[22]  Risk models with dependence between claim occurrences and severities for Atlantic hurricanes , 2014 .

[23]  R. Ambagaspitiya,et al.  On the distributions of two classes of correlated aggregate claims , 1999 .

[24]  W. Zucchini,et al.  Hidden Markov Models for Time Series: An Introduction Using R , 2009 .

[25]  Junyi Guo,et al.  On a correlated aggregate claims model with Poisson and Erlang risk processes , 2002 .

[26]  Bernard Wong,et al.  A Micro-Level Claim Count Model with Overdispersion and Reporting Delays , 2015 .

[27]  Roberta Paroli,et al.  Poisson Hidden Markov models for time series of overdispersed insurance counts , 2000 .