Algorithm Predicts Personal Life Events With High Accuracy, Including Likelihood of Earlier Death
• Danish researchers used machine learning to predict personal life events and outcomes, including predicting if someone will die within a few years • The algorithm, called life2vec, was 78% accurate at predicting if someone would die after 2016 when tested on a dataset of 100,000 people • Life2vec outperformed other models by 11% in predicting mortality • Factors that made earlier death more likely included being male, having mental health issues, and being a skilled worker rather than a manager • Limitations include a narrow timeframe and questions around applying the model outside of Denmark, but experts say more advanced versions will emerge soon