Key attribute mining method based on attention mechanism

Abstract

The invention discloses a key attribute mining method based on an attention mechanism, and belongs to the technical field of data mining. The key attribute mining method based on the attention mechanism comprises the steps that preliminary attribute clustering is conducted on data based on expert experience, the attention mechanism based on a multi-layer perceptron is constructed for the preprocessed data, and a combat result prediction model and an attention mechanism model are combined; and the weight distribution output based on the attention mechanism model is multiplied by the feature vector of each attribute in the prediction model to obtain more accurate feature representation for final combat result prediction, and the weight distribution output based on the attention mechanism model is optimized through the final combat result. Key attribute information is extracted by using an attention mechanism, and interference of redundant attributes is reduced, so that combat result prediction with high accuracy is realized, and a commander is better helped to master battlefield situation changes macroscopically.

Description

technical field [0001] The invention relates to a key attribute mining method based on an attention mechanism, and belongs to the technical field of data mining. Background technique [0002] In the information age, non-contact warfare and network-centric warfare have become the main forms of future warfare. The combat process is manifested as a comprehensive and complex confrontation process in which various combat forces under the control of multiple mechanisms act on multiple related fields. Combat forces are composed of " The development of "simple structure" to "complex structure", and the complexity and asymmetry of war have put forward higher requirements for the development of new combat forces. In modern warfare, commanders need real-time assistance from some combat simulation systems in order to face these complex battlefield situations in real combat scenarios. Therefore, in order to solve this problem, it is necessary to dig out the key factors in the current sit...

AI Technical Summary

Problems solved by technology

[0005] Based on the above research status, the key attribute mining method for multi-dimensional data mainly has the following problems: First, the attribute extraction method based on classification or clustering is not suitable for the deep-level inter-attribute in the complex and dynamic environment of military command and control. The relationship understanding may not be in place, and the model generalization ability is not good

Second, based on the method of association rules, it is not easy to extract useful rules in combat scenarios with complex rules

Images