Vector space computing strength predicting method and system based on fully random forest

Abstract

The invention discloses a vector space computing strength predicting method and system based on a fully random forest. By inputting all features relevant to the vector space computing strength, multiple full regression trees are trained, computing strength modeling for a vector space computing domain with various features is achieved, the prediction result of the fully random forest is optimized,prediction values which are distinctly different from the prediction result are removed, the prediction precision of the fully random forest is improved, and the vector space computing strength is precisely predicted in a parallel computing environment. By means of the vector space computing strength predicting method and system based on the fully random forest, in the training process of the fully random forest, the training sample of each regression decision-making tree is randomly selected from an original sample, the selected features comprise all features of the original sample, and the model can adapt to prediction of the vector space computing strength with few important features and many redundancy features. A basis can be provided for parallel computing resource balanced scheduling and allocating, and the parallel computing efficiency is improved.

Description

technical field [0001] The invention relates to a high-performance geospatial computing technology, in particular to a vector space computing strength prediction method and system based on a complete random forest. Background technique [0002] The modeling and prediction of vector space computing intensity is one of the key issues to be solved in high-performance spatial computing. In a parallel computing environment, parallel computing resources can be scheduled in a balanced manner according to the magnitude of vector data space computing intensity. It is beneficial to obtain a higher speed-up ratio. However, the existing vector space calculation intensity prediction model is mainly based on the linear function model, which considers some vector space calculation intensity features, but does not consider the influence of other vector space calculation intensity features, and cannot deal with the interaction between different features. More importantly, even with the exce...

AI Technical Summary

Problems solved by technology

However, the existing vector space calculation intensity prediction model is mainly based on the linear function model, which considers some vector space calculation intensity features, but does not consider the influence of other vector space calculation intensity features, and cannot deal with the interaction between different features.

More importantly, even with the excellent random forest modeling method, it cannot adapt to the accurate prediction of vector space calculation intensity with few important features and many redundant features, making it difficult to achieve accurate measurement of vector space calculation intensity

Images