Research on time factor prediction in the Lee-Carter model—the combined model based on Lp-norm
For longevity risk, this paper focuses on accurately predicting mortality rates and proposes an improved time-factor prediction model to enhance the accuracy of mortality rate forecasting. Traditional time-factor model exhibits limitations in handling the nonlinearity and long memory characteristics...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis
2025-12-01
|
| Series: | Research in Statistics |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/27684520.2025.2528353 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | For longevity risk, this paper focuses on accurately predicting mortality rates and proposes an improved time-factor prediction model to enhance the accuracy of mortality rate forecasting. Traditional time-factor model exhibits limitations in handling the nonlinearity and long memory characteristics. To overcome this, two combination models are constructed: the series model (CPGL) and parallel model (BPGL), which utilize the Lp-norm method to integrate the ARIMA, GALSSVM, and LSTM model, optimizing the weights of different models to improve predictive performance. Empirical analysis uses male mortality rate data from the UK, USA, Taiwan, Hong Kong, and mainland China. Four evaluation metrics are employed to compare the predictive performance of the CPGL, BPGL, and ARIMA models. The results demonstrate that the combination models significantly outperform the ARIMA model in terms of time-factor prediction accuracy, with BPGL showing the best performance. Additionally, a comparison of mortality rate prediction accuracy from both age and time perspectives further confirms the superiority of the combination models over the ARIMA model. Thus, the proposed method, which combines the ARIMA model with other models, significantly enhances the accuracy of time-factor prediction and provides more precise mortality rate forecasts. This offers new insights and scientific evidence for research in longevity risk. |
|---|---|
| ISSN: | 2768-4520 |