Method, apparatus, device and machine-readable medium for detecting abnormal data using an autoencoder

An apparatus for detecting abnormal data, the apparatus comprising:

• a creating module configured to use at least two items of data to create at least two detection data sets, wherein
• the at least two items of data are sequentially collected and mutually associated, and
• each detection data set of the at least two detection data sets includes a first number of items of sequentially collected data in the at least two items of data;
• a processing module configured to use an autoencoder to process the at least two detection data sets, and to output result data sets respectively corresponding to the at least two detection data sets, wherein
• the first number is equal to a number of neurons in an input layer of the autoencoder, and
• autoencoder is trained using training data having a regular pattern of variation identical to that of the at least two items of data;
• and
• a determining module configured to determine, as the abnormal data, data not having the regular pattern of variation in the at least two items of data, based upon the at least two detection data sets and the result data sets corresponding thereto, the determining module including,
• a computing module configured to use a detection data set containing an item of data in the at least two items of data, and a result data set corresponding to the detection data set, to compute a residual error of the item of data, to obtain respective residual errors of the at least two items of data, and
• a searching module configured to identify, as the abnormal data, data with a residual error greater than a first threshold in the at least two items of data.

A device for detecting abnormal data, the device comprising:

• at least one processor;
• and
• at least one memory, storing an executable instruction which, when executed, causes the at least one processor to,
• create at least two detection data sets, the at least two detection data sets being created using at least two items of data, the at least two items of data being sequentially collected and mutually associated, and each detection data set of the at least two detection data sets including a first number of items of sequentially collected data in the at least two items of data,
• process, using an autoencoder, the at least two detection data sets to generate result data sets respectively corresponding to the at least two detection data sets, the first number being equal to a number of neurons in an input layer of the autoencoder, and the autoencoder being trained using training data having a regular pattern of variation identical to the at least two items of data;
• and
• determine, as the abnormal data, data not having the regular pattern of variation in the at least two items of data, based upon the at least two detection data sets,
• wherein the executable instruction, when executed, causes the at least one processor to determine the data as the abnormal data by
• computing, using a detection data set containing an item of data in the at least two items of data, and a result data set corresponding to the detection data set, a residual error of the item of data, to obtain a respective residual error of the at least two items of data, and
• identifying, as the abnormal data, data with a residual error greater than a first threshold in the at least two items of data.