Distributions-free Martingale test Distributions-shift for Swarm Behavior Prediction
Zepu Xi · 2022 International Joint Conference on Neural Networks (IJCNN) · 2022
The timely detection of swarm behavior is critical to ensure a reliable and appropriate response to update any behavior prediction to complex swarm phenomena. A standard assumption under the method of machine learning is that the observed data are generated from a fixed but unknown distribution. However, when solving practical swarm behavior prediction problems it is remarked that the data points are observed batch by batch. So in this paper, we are devoted to testing the assumption of distributions-shift online: the observed data arrive one by one, and after receiving each object, the machine learning algorithms give a prediction label, we would like to have a valid measure of the degree to which the evidence to against the assumption of the fixed distribution. We report the experimental performance of distributions-free martingales testing on the swarm benchmark data set, the results show a bona fide fact that the distributions-shift detection testing is an inescapable reality for the original order data set.