摘要
由于我国税收收入存在高度的非线性、耦合性和多因素的影响,故而对其进行预测是传统的预测方法难以胜任的。首先,提出偏最小二乘支持向量回归对我国税收收入进行预测的思路。其次,由于参数集(C,σ2)直接影响支持向量技术的预测优劣,故采用改进的遗传算法对参数集进行全局寻优,这样既保证了处理非线性的优势,又确保了支持向量回归模型的稳定性与精确性。结果表明,预测精度有着显著提高,说明了该模型的有效性与实用性。
Chinese tax revenue is non - linear and coupled, and is influenced by many factors. There fore, traditional forecasting methods are not sufficient to predict the value of it. In this paper, partial least square support vector machine （PLS - SVR） is used to construct a tax revenue prediction model. In order to solve the problems, a financial time series forecas- ting model based on Genetic Algorithm which is used to optimize parameters of SVR has established. By doing so, this model can deal with the nonlinearity and multi - factors of tax revenue, and ensure stability and accuracy of support vector machine based regression. Case study on Chinese tax revenue during the last 30 years demonstrates that the optimized PLS - SVR model is much more accurate than other prediction methods.
出处
《科技管理研究》
CSSCI
北大核心
2014年第11期197-200,共4页
Science and Technology Management Research
基金
国家自然科学基金项目"不确定信息环境下基于大规模数据的趋势预测和智能决策方法究"(71101041)
"云计算环境下智能决策方法研究"(71071045)
