THE CHANGE OF E-INSURANCE IN TURKEY AND FORECASTING E-INSURANCE PREMIUM PRODUCTION IN LIFE INSURANCE BY NONLINEAR AUTOREGRESSIVE NEURAL NETWORKS

Abstract

E-commerce, which can be defined as trade through electronic systems, is also emerging in the insurance sector. The literature refers to sales through e-commerce as e-insurance in sales by type Therefore, e-insurance has become important in the global insurance market. This study aim to investigate the change in e-insurance in Turkey and forecast premium production in e-insurance for life insurance. Non-linear autoregressive neural network used for time series analysis was applied for the forecasting. This quantitative research draws its data from a secondary source of premium tables on sales by type released by the Insurance Association of Turkey. The descriptive findings indicate that as total policy premium from e-insurance was 34.151.563 with 0,1% of the market in 2015, it was 2.659.041.678 with 2,5% of the market in 2021. The forecasting of life insurance was analyzed using non-linear autoregressive neural network by e-insurance premium values of 72 monthly for the period of 2016:01-2021:12. The most appropriate network model with (12-5-1) architecture structure and learning with back propagation algorithm based on the Levenberg-Marquardt algorithm was established. After the training model, the mean squared error selected as a performance criterion is the 2,07982e-4, and the regression determination coefficient is also R²= 0,93. The forecast obtained using the network model shows that e-insurance tends to increase for years to come. As a result, e-insurance is likely to gain significant market share.