Industrial control system anomaly detection method based on Taguchi method

The invention provides an industrial control system anomaly detection method based on a Taguchi method, and the method comprises the following steps: carrying out the oversampling and normalization processing of a data set, and dividing the data set into a training set and a test set; carrying out binarization processing on the GSK algorithm by using a U-shaped transfer function; aiming at a specific target problem, customizing a group of optimal parameter combinations for the binarized GSK algorithm by using a Taguchi algorithm, and providing parameter settings according to the property of a specific target; selecting features through a customized binary GSK algorithm; and inputting the selected optimal feature subset into an LSTM neural network to train a model, and obtaining an LSTM-based industrial control system anomaly detection model. According to the technical scheme, the problems that in the prior art, an anomaly detection model is prone to being affected by redundant features, the detection precisio