Human Death Will Be Predicted By AI Based On Data Of 500 Thousand People

Scientists have built a computer system based on artificial intelligence that is capable of predicting the risk of premature death due to chronic diseases that often occur in middle-aged people.

Humans are always curious about their future, so many individuals have searched on Google for their death prediction. A lot of websites have been built up to answer the question such as deathtimer.com, death-clock.org, or fatefulday.ei. Most of these sites offer scientifically unconvincing results. It only collects basic information such as smoking intensity, or BMI, ...

Now, advances in the field of health analysis have helped scientists come up with answers to this difficult question. Thanks to the powerful help of machine learning algorithms, humans are able to know their expected death time in the future.

Researchers at Nottingham University in the United Kingdom have taken the health data of more than half a million people between the ages of 40 and 69 in U.K. BioBank. This is long-term research on the relationship between surrounding factors and genetic predisposition. They are included in an artificial intelligence system, resulting in far more accurate results than the standard model developed by humans and currently used in predicting death time due to chronic disease. This study was published in the PLOS ONE special edition on Wednesday. Dr. Stephen Weng, a lecturer in data science and epidemiology at the University of Nottingham, and researchers in the group used the new algorithm and produced nearly 79% of the early death times of some case.

Although there are still some notable failures, the development focus in preventive health care has continued to increase in recent years. People not only go to private clinics to control their health but also care about insurance service. They always try to cut down on health care costs for each individual. When vegetables, fruits, and turnips are included in prescriptions, they help reducing health care costs by up to $40 billion per year, according to a recent study on PLOS ONE.

This is not the first time Weng and his research team have developed a program to predict human health. In 2017, a machine learning algorithm was developed by this group to predict stroke and heart attacks through information on electronic health records. This model gave 8% better results than the cardiovascular risk prediction method used by doctors.

The success of predicting the death of chronic disease has brought motivation for researchers to continue developing this model. They hope that it can predict the death caused by many other diseases. They tracked and analyzed biometric data, demographics, medical factors and lifestyles of people involved in the research process. After that, the data will be provided to two types of machine learning as "random forest" and "deep learning".

Weng said that the test models are based on the principle of “trial and error”, so having to provide and explain a large amount of data to be able to give the right answer. The "random forest" will be based on the risk factors in order to determine whether death will occur soon. Meanwhile, "deep learning" has the ability to gather knowledge like neural networks in human brains, then it filters and maximizes models.

Weng said the human doctor's risk factor determination will be based on previous evidence or signs, but machine learning algorithms will make decisions about important risk factors through the system itself.

There are many algorithms in this work that are too complex and difficult for scientists to analyze, it is hard for them to how the system has evolved. So that the work of replicating this finding is still a challenge, according to Stephen Weng. Scientists have also published research details, including methods, all code, and results to the community.

In fact, AI has successfully completed its work in human health care. But Weng wishes that people will use it as a tool to help the jobs be completed more quickly, more accurately, not as a complete replacement of the human role. In particular, people should not oppose the emergence of artificial intelligence in the medical field.