Classification Algorithms In Empirical Analysis Of The Model Tax Assessment

Tax assessment refers to the taxation entity's use of fiscal and taxation information to carry out an assessment of the tax-related actions of the object of supervision.The third phase of the Golden Tax Project has strengthened the implementation of the assessment of tax risk for big data,and the informationization of the tax assessment system has been fully enhanced.The taxation management in the era of big data needs to cross-analyze taxation data of enterprises with artificial intelligence technology and data mining algorithms to monitor the tax-related behaviors of enterprises in real time.This paper integrates the tax assessment system and machine learning to create a more intelligent classification algorithm.First of all,it discusses the connotation and characteristics of tax assessment and the current academic research on this topic.Secondly,it introduces five kinds of classification algorithms and theoretical knowledge in machine learning,and explains the goodness and badness of each classification algorithm and its applicable scenario.Then,for data on property taxation in Chongqing enterprises,we preprocessed and cleaned the data,and then selected important indicators using feature engineering.This paper presents an empirical analysis of the random forest,GBDT,XGBOOST,and LightGBM algorithms.Based on the LightGBM algorithm,a LightGBM and Logistic hybrid classification algorithm is proposed,and the LightGBM algorithm is improved.Finally,compare the results of the five algorithms and analyze the causes.The research in this paper shows that the XGBOOST algorithm has the best classification effect in the corporate property tax assessment data,and the GBDT algorithm classification effect is the second.