Back

Hybrid Deep Learning and Lee-Carter Model for Mortality Forecasting: A Study of US Adults Aged 35-80

Abaitey, C.; Abakah, E.; Asare, C.; Oppong, E.; Baba-Sadiq, F. M.; Appiah, E.

2025-10-15 bioinformatics
10.1101/2025.10.13.682203 bioRxiv
Show abstract

This study introduces a hybrid approach that enhances mortality forecasts by integrating machine learning techniques specifically Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), and Neural Networks (NN) with the traditional Lee-Carter model. After deriving the time index{kappa} t from the Lee-Carter model, these deep learning models were employed to capture complex temporal patterns, which were then incorporated into the Lee-Carter frame-work to improve forecasting accuracy. The hybrid models were evaluated using historical mortality data from the United States, covering ages 35 to 80 years from 1975 to 2020. Among the models tested, the LSTM model outperformed all others, demonstrating superior capability in capturing time dependencies and producing more accurate mortality forecasts. The integration of LSTM with the Lee-Carter framework led to significant improvements in predictive accuracy. This research demonstrates that combining traditional statistical approaches with modern deep learning techniques, particularly LSTM, offers a powerful method for enhancing mortality forecasting by effectively modeling time-dependent patterns. These findings provide valuable insights and tools for policymakers, actuaries, and healthcare professionals to improve planning and decision-making.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.