Multi-sensor abnormal value detection method and device, computer equipment and storage medium

Abstract

The embodiment of the invention discloses a multi-sensor abnormal value detection method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a multivariate original signal; training a first auto-encoder by using a reconstruction error according to the multivariate original signal, and performing VMD decomposition on a hidden layer in the first auto-encoder to obtain a decomposed signal; sampling data features of the decomposition signal in a bagging mode to obtain a feature value; inputting the feature value into a second auto-encoder and optimizing two auto-encoders in the second auto-encoder to determine an upper bound estimated value of the reconstructed feature and a lower bound estimated value of the reconstructed feature; determining a sample label corresponding to the feature value according to the decomposition signal, the upper bound estimation value of the reconstruction feature and the lower bound estimation value of the reconstruction feature; and determining an abnormal value according to the sample label corresponding to the feature value. By implementing the method provided by the embodiment of the invention, the requirements of large-scale data can be met, and the robustness of the detection algorithm is good.

Description

technical field [0001] The present invention relates to an outlier detection method, more specifically to a multi-sensor outlier detection method, device, computer equipment and storage medium. Background technique [0002] An autoencoder is a neural network trained by unsupervised learning that is trained to learn a reconstruction close to its original input. An autoencoder consists of two parts, an encoder and a decoder. A neural network with a single hidden layer has Z = σ(W 1 *Y+b 1 ) and Y * =σ(W 2 *Z+b 2 ), W and b are the weights and biases of the neural network, and σ is the nonlinear transformation function. The optimization goal of AE||Y—Y * || min ; Z=σ(W 1 *Y+b 1 The encoder in ) maps the input vector Y to the hidden space Z by affine mapping after nonlinearity. Y * =σ(W 2 *Z+b 2 The decoder in ) maps the hidden representation h back to the original input space as the reconstructed signal through the same transformation as the encoder. Original inp...

AI Technical Summary

Problems solved by technology

[0004] However, for sensor data, as the detection time increases, the data size increases exponentially, the current outlier detection methods cannot meet the needs of processing large-scale data, and the robustness of the current algorithm is not good

Images