System, index optimization method and index optimization device

Abstract

The invention provides a system, an index optimization method and an index optimization device. The method comprises steps: all already-used index parameters and numerical values thereof of the system, and all to-be-selected index parameters and numerical values thereof are acquired; based on a preset data dimension reduction algorithm, dimension reduction processing is carried out on all already-used index parameters and the numerical values thereof, and a corresponding feature parameter set and the numerical values thereof are acquired; with the numerical values of the feature parameter set as input and the numerical values of all to-be-selected index parameters as target output, a preset machine learning model is trained, and prediction values of the numerical values of all to-be-selected index parameters are acquired; a difference measurement value between the numerical value of each of the all to-be-selected index parameters and a corresponding prediction value is acquired; and a preset number of to-be-selected indexes with the maximum difference measurement values are selected as newly-added indexes of the system. The index optimization efficiency can be improved, and system performance jitter is reduced.

Description

Technical field [0001] This application relates to the field of data processing technology, and in particular to a system and its index optimization method and device. Background technique [0002] As time goes by, in the actual application process of some systems, the relevant statistical information is continuously accumulated and enriched. Through the analysis and processing of the statistical information, it may be found that new indicators need to be added and the system reconstructed on this basis To improve its performance. [0003] With the continuous development and change of computer networks and information technology, there are currently some systems that have the function of automatically constructing new index sets. These new index sets can help the system adapt to new changes, thereby helping to improve system performance. However, because the number of new indicators in the new indicator set is usually large, and some of the systems (such as online systems) have li...

AI Technical Summary

Problems solved by technology

[0005] However, the inventors of the present application found that the above method needs to traverse each new index in the new index set one by one, which is very time-consuming

At the same time, if the existing system is already relatively complex, adding a single indicator may not actually bring actual performance improvement to the system

Sometimes, the jitter of system performance may even be caused by the selection of random parameters. Therefore, in general, it is often necessary to add a set of indicators at the same time to see the actual effect.

According to the existing technology, the traversal complexity of selecting a set of indicators from a new indicator set is exponential, which will occupy too many system resources, and it is almost difficult to implement in engineering

Images