You might have misunderstood the point of the ML strategy. The idea is to train a model on explainable event data such as weather balloons or wildlife as collec

You might have misunderstood the point of the ML strategy. The idea is to train a model on explainable event data such as weather balloons or wildlife as collected via various existing sensors. Then take this model and use it on historical sensor data to identify events that do not match the traits of the known events. I feel like this is a pretty reasonable use for ML.