第三届挑战赛B1-基于数据挖掘技术的市财政收入分析预测模型

全国大学生数据挖掘竞赛 优秀作品 作品名称：基于数据挖掘技术的市财政收入分析预测模型 荣获奖项：一等奖 作品单位：长江大学 作品成员：吴海明陈育方青 指导教师：胡中波 基于组合预测模型的广州市2015年地方财政收入预测 摘要：从广州市“十二五”规划纲要中，可以看出广州市作为我国一直以来的财政大省，目前正处于经济转型升级的关键时期，准确地预测地方财政收入对保障广州经济转型升级的成功进行有极其重大的意义。合理的预测模型能给出更为精准的预测值。本文基于广州市统计年鉴数据，通过由灰色系统中的灰色关联度算法、GM(1，1)预测模型和BP神经网络预测模型组成组合预测模型对广州2015年的地方财政收入进行预测。然后，基于数据挖掘的结果，给广州市财政局提出一些财政建议。具体分析过程如下： 第一步：基于广州统计信息网统计年鉴，从经济学角度考虑地方财政收入的影响指标。主要是从两个方面着手考虑：一是地方财政收入的组成，地方财政收入主要由公共地方财政收入和政府性基金收入组成；二是地方财政收入的规模，影响地方财政收入规模的因素主要经济发展水平，分配制度和分配政策，还有物价水平，而对经济发展水平的研究通常归结于对失业率，通货膨胀还有经济增长率的研究，另一方面是经济结构，即产业结构的变动也是经济发展水平的一大影响因素。对于上述影响指标，通过灰色关联度分析算法，筛选出影响地方财政收入的主要因素。 第二步：基于第一步分析的主要影响因素，通过组合模型进行财政预测。 1．基于主要影响因素的数据，通过GM(1，1)预测模型，对各个因素的原始数据分别进行GM(1，1)预测，得出1999-2015年的各个因素的预测值。 2．基于GM(1，1)预测模型的预测值，通过BP神经网络预测方法，把GM(1，1)预测的2014-2015年的各个因素的预测值代入BP神经网络进行预测，得出2015年的广州市地方财政收入的预测值。 第三步：在预测地方财政收入的过程中，了解到影响广州市地方财政收入的主要因素，一方面我们将结合得出的地方财政收入预测值和主要因素向广州市政府给出建议，进而实现广州市2015年的地方财政收入目标。另一方面，我们分析了广州市财政支出的重点支出方向，从财政支出的角度给出建议。 本文通过以上三步的研究，确定了影响广州市地方财政收入的十个主要影响因素：地方财政收入、增值税、营业税、企业所得税、个人所得税、城市维护建设税、契税、行政性收费收入、附:上级补助收入、消费税和增值税税收返还、 所得税基数返还；预测出了广州市2014、2015年地方财政收入分别为：24516007万元、27930804万元；基于分析和预测过程，从优化税制、完善地方税种，规范税收管理，推进产业结构的优化和升级三个方面给出了合理的财政建议。 关键词：地方财政收入预测；灰色关联度算法；GM(1，1)预测模型；BP神经网络；财政建议 Theforecastoffinancialrevenueofguangzhouin2015basedoncombinationforecastingmodel Abstract:Accordingtotheessentialsofthe12thfive-yearplan,wecanknowGuangzhou,whosefinanceremainsaleadingpositioninourcountry,isatacriticalstagewhenGuangzhouistransformingandupdatingitseconomy,soit’ssignificantmeaningforGuangzhoutoforecastthefinancialrevenueofGuangzhouin2015.Morepredictedvaluecanbeobtainedbyrationalforecastingmodel.Basedonstatisticyearbookdataofguangzhou,thearticlegetpredictedvalueoffinancialrevenueofguangzhouin2015bycombininggrayrelationaldegreeanalysisalgorithm,GM(1,1),andBPneutralnetworkmodeltogether.Then,wegivesomefinancialadvicetothegovernmentbasedonpredictedvalue.Theprocessofconcreteanalysis is as follows. Thefirst step, based on statistic yearbook data of guangzhou, we analyzeinfluential factors of financial revenue from the point of economics, mainly two points,one is the component of financial revenue, which is comprised of public financialrevenue and revenue from government-controlled .another is the scale of financialrevenue,whose influential factor areeconomic development level, distribution systemand distribution policy and price level, and the study of economic development levelis usually completed by not only studying unemployment rate, inflation and economicgrowth rate, but also economic structure, namely industrial structure .For the impactindicators, through gray relational degree analysis algorithm,we selectthe main factorsaffecting fiscal revenue. The second step: Based on the first analysis ofthe main factors, financial revenueis completed through the combination of models. First,based on the main factors of the data, through GM (1, 1), the original dataof every factor will befiltrated by GM (1, 1) forecasting and the forecast values ofeveryfactors.during 1999-2015 will be obtained . Second, based on predictive value of GM (1, 1), GM (1, 1) predictive value ofthe factors in 2014-2015 are put into BP neural network to predict, then predictivevalue of guangzhou financial revenue in 2015 will be gotten. The three step: the main influential factors of Guangzhoufinancial revenue havebeenknown in the process of predicting financial revenue.On the onehand,wewillgiveoursuggestiontogovernmentofguangzhoubasedonthepredictivevalueoffinancialrevenuewehaveobtainedso as to realize Guangzhou fiscal revenue target of2015. On the other hand,not only will we analyze the key expenditure direction of thefinancial expenditure ofGuangzhou,but we also come up with the suggestion from thepoint of fiscal expenditure. Conclusion, by above three steps study, ten main influential factors of financialrevenueofguangzhouhavebeenconfirmed,whicharelocalfinancialrevenue,added–valuetax,businesstax,enterprise income tax,individualincometax,urban maintenance and construction tax,deed tax,administrative feeincome, grant from the higher authority, consumption tax and added- value tax returnsand base number of income tax returns and local financial revenue of guangzhou in2014and 2015 respectively are 253.16007 billion yuan and 289.30804 billion yuanhave been forecasted .Based on the process of analysis and forecast, suggestions aregiven from three aspect, optimizing tax system and completing local tax,regulate taxmanagement, boosting optimization of industrial structure and upgrade . Keywords:the forecast of financial revenue,gray relational degree analysisalgorithm, GM(1,1), BP neutral network model, financial advice. 目录 一．挖掘目标........................................................................................................1二．分析方法、过程和总结................................................................................22.1总体流程..................................................................................................22.2具体步