Many natural disasters offer limited or no advance warning of the destruction to come, and predictions can vary as the event progresses. With lead times being limited, critical resources can often be out of place or unavailable, putting additional strain on both those directly affected by a disaster, as well as those attempting to respond and provide aid.
To help improve planning and responses, considerations are being made regarding the deployment of machine learning, according to an article in Scientific American written by Seth Guikema, a professor in the Department of Industrial and Operations Engineering and the Department of Civil and Environmental Engineering at the University of Michigan. Taking advantage of the ability to churn through large volumes of data on previous disasters and responses to make predictions about future events and trends, machine learning could assist in ensuring resources are deployed efficiently at both the local and regional level, or to identify likely future failures or secondary effects from disasters.
The article also notes the limitations of machine learning: namely, the parameters set by the quality and quantity of information provided. Should the data set lack information on specific types of disaster, or should a new type of disaster occur, machine learning will be unable to predict it, according to Guikema. Additionally, the outputs from machine learning systems still require human judgement to interpret and implement them. “Machine learning cannot and should not replace traditional methods of disaster response. Expert human judgement is absolutely critical, given the complexity and magnitude of the situations,” writes Guikema.
