Perceptron Model of Forecasting Life Exapectancy via Insurance Lee-Carter Mortality Function

Forecasting of mortality function is important for many field of human work like insurance companies, government projections of the human assets, medical research. During past years many models were presented. Most common Lee-Carter model is based on the log function on mortality rate which includes as input variables age, year of mortality function and bias, which also enables predicting the life expectancy. In this paper a perceptron based model with minimum number of nodes in the network having custom transfer function is proposed. Results are compared with Lee-Carter and other neural network based models by using MSE type of error. This model is simpler than other neural networks and is easier to handle adjusting the weights while computing results are rather comparable with those of more complex neural network models. © 2018 IEEE.*G. M. Dimirovski was supported by the Dogus University Fond for Science