Prediction Of Computational Advertising Conversion Rate Based On Factor Space

With the development of the field of computing information,more and more researchers pay attention to the prediction technology of computational advertising conversion rate.At the same time,the factor space theory is developing in the field of intelligent information analysis and processing.Therefore,the prediction of computational advertising conversion rate based on factor space theory is a deep and meaningful research.This paper focuses on the problem of feature extraction and hybrid stacked prediction model of computational advertising conversion rate.The main achievements are as follows:Firstly,This paper introduces the basic information and data cleaning of calculating advertising conversion rate,analyzes the data characteristics,and eliminates invalid or weak expression data.Through data visualization analysis,the selection range of later feature combination mode is reduced.Secondly,In order to solve the problem of how to enhance the expression of features,the feature extraction method based on factor space theory is applied to the prediction of advertising conversion rate for the first time.Five data features are selected to do the experiment,and the effective factors are extracted successfully.Then through the feature importance ranking,it is found that the new features have stronger feature expressiveness than the original features.Thirdly,Aiming at the problem of complex target optimization in feature extraction method based on factor space theory,this paper proposes a factor genetic algorithm for complex fitness function.This paper mainly uses the factor space theory and TOPSIS idea to add the chromosome with the most genetic properties,and then put it into the multi population genetic mechanism.Experiments show that in the environment of complex fitness,the factor genetic algorithm can achieve the optimal iterative effect faster than the multi population genetic algorithm.Finally,Aiming at the problem of selecting the prediction model for calculating advertisement conversion rate,this paper introduces the supervised learning mechanism,and compares the advantages and disadvantages of XGB and LGB algorithms.The logarithm loss value of 0.0886468 is calculated by stacking and mixing XGB and LGB as the conversion prediction model,and the mixed weight selection problem of the model is solved by using factor genetic algorithm.