Classifier chain local detection and mining algorithm

Abstract

The invention discloses a classifier chain local detection and mining algorithm, and the algorithm comprises the steps: optimizing a classifier chain, used for solving a big-data flow mining problem, through employing a local learning algorithm, and simulating a learning problem of the classifier chain through employing a multi-user multi-arm problem with finite feedback. The proposed algorithm employs a cooperative and distributed method for learning, so the algorithm can select an optimal classification structure and learns the effect of a classifier in real time. The data processing process just needs one path: a data set, thereby enabling the processing delay and the internal memory demands of the processor to be minimized. In addition, the proposed algorithm does not need to employ a learning process of a central unit in the distributed classifier for operation cooperation, so the algorithm reduces the demands of exchange between various types of classifiers. Meanwhile, the proposed algorithm carries out the learning according to a comprehensive task mining effect, not according to the subtask effect, thereby reducing a large amount of feedback information.

Description

technical field [0001] The invention relates to the field of local learning algorithms, in particular to a classifier chain local detection and mining algorithm. Background technique [0002] In recent years, big data has been widely used in many fields, such as social media analysis, video surveillance, network security monitoring, etc., all of which require the analysis and processing of raw data streams to obtain real-time valuable information. Existing methods for dealing with data stream mining problems with limited resources all rely on load shedding, and decide the discarding strategy according to the given data characteristics. Since the discarded data will also play an important role in the next classifier, the load shedding method generally cannot obtain the best effect of point-to-point. Existing methods assume that the performance of classifiers is known, and require effective information exchange between classifiers, which is often impossible to achieve in prac...

AI Technical Summary

Problems solved by technology

Since the discarded data will also play an important role in the next classifier, the load shedding method generally cannot obtain the best effect of point-to-point

Existing methods assume that the performance of classifiers is known and require effective information exchange between classifiers, which is often impossible to achieve in practical applications, and these methods cannot perform distributed processing

Images