【摘要】 目前进行纳税评估和预测工作主要依赖于纳税评估人员的人工判别和分析,这样导致税务评估人员工作量较大,而且所得的评估结果也不准确。为了解决这一问题,提出了基于Adaboost-PSO-SVM的纳税评估模型。利用PSO优化SVM弱分类器,再用Adaboost将多个PSO-SVM组合成为强分类器进行纳税评估。实验结果表明在纳税评估方面,相比于单个SVM弱分类器,Adaboost-PSO-SVM强分类器的准确率由94%提高到了99%。在纳税评估的基础上,利用SVM回归机实现对纳税数据变化趋势和变化空间的预测,结果表明包含纳税评估结果的预测模型的预测效果更好。更多还原

【Abstract】 At present,the tax assessment and prediction work is mainly dependent on the artificial judgment and analysis of tax assessment personnel,which bring a large workload to tax assessment personnel and the evaluation results are not accurate.In order to solve this problem,this presents a tax assessment model based on Adaboost-PSOSVM.It use PSO to optimize SVM weak classifier,and then utilize Adaboost to combine multiple PSO-SVM into a strong classifier for tax assessment.Experimental results show that,compared with single SVM weak classifier,the accuracy of Adaboost-PSO-SVM strong classifier is increased from 94% to 99%.On the basis of tax assessment,we apply SVM regression machine to realize the prediction of the change trend and range of variety of tax data,the results show that the prediction model contains tax assessment is better.更多还原