Forecasting method and system based on forest discrimination model

Abstract

The invention discloses a prediction method and system based on a forest discrimination model. The method includes: step 1, using modeling data to model based on a random forest algorithm, binning the modeling data, and passing the discrimination degree Solving the model to obtain a modeling result; step 2, scoring the data to be predicted according to the established model and the modeling result to obtain a prediction result. The present invention can provide a simple, expanded, and evidence-based binning method, which simplifies the consideration of data, and can use this method to obtain better binning without the need to be proficient in advanced calculation theory As a result, better prediction results can be obtained. This also improves the efficiency of modeling and the accuracy of the model to a large extent.

Description

technical field [0001] The invention relates to the technical field of scorecard modeling, in particular to a prediction method and system based on a forest discrimination model. Background technique [0002] The scorecard is a statistics-based analysis tool, which can predict the probability of a specific result in the future by analyzing the phenomenon that has occurred, based on historical data and various related factors. The basis of the analysis is a series of interrelated factors, variables, predictive values ​​or independent variables of the regression equation that describe the historical data, which we call predictor variables. A specific outcome to be predicted is called the target variable. Scorecard modeling technology is based on the scorecard to establish a mapping relationship between predictor variables and target variables. This mapping relationship is the scorecard model. The commonly used model methods of the scorecard model include linear regression, l...

AI Technical Summary

Problems solved by technology

[0012] At present, data analysts have two main problems in the variable binning process: one is the low degree of automation, which takes a lot of time, and the binning results are highly dependent on the analyst's experience, which is not conducive to obtaining good binning results. Mature technical guarantee; second, the variable binning process only targets a single variable, without considering the relationship between multiple variables

Therefore, the existing methods do not take into account the correlation between different variables, that is, when dividing the existing samples, the factors of multiple variables are not considered comprehensively.

Images