資源描述:
《金融數(shù)據(jù)挖掘.doc》由會員上傳分享,免費在線閱讀,更多相關內容在行業(yè)資料-天天文庫。
1、《金融數(shù)據(jù)挖掘與應用》課程作業(yè)基于GLM(廣義線性模型)的數(shù)據(jù)分析SAS里的GLM應用在實際中比較廣泛,對數(shù)據(jù)的分析具有比較強的普適性。趨勢面回歸分析(TrendAnalysis)是以多元回歸分析為理論基礎的一種預測與統(tǒng)計技術。它用空間坐標法進行多項式回歸,從中估計出最佳的回歸模型,因此也被稱為趨勢面分析,當不知道手中的數(shù)據(jù)呈線性還是非線性相關時,可以采用趨勢面數(shù)據(jù)分析方法,以便找出擬合數(shù)據(jù)的最佳統(tǒng)計預測模型。本文運用GLM對一定的數(shù)據(jù)進行GLM分析。一、數(shù)據(jù)與要求此處選取15名吧不同程度的煙民的每日飲酒
2、(啤酒)量與心電圖指標(zb)的對應數(shù)據(jù)。然后設法建立zb與日抽煙量(X)/支和日飲酒量(y)/升之間的關系。序號組別日抽煙量(x)/支日飲酒量(y)/升心電圖指標(zb)113010280212511260313513330414014400514514410622012270721811210822512280922513300102231329011340144101234515420133481642514350184501535519470二、運用GLM過程進行趨勢面分析1.趨勢分析的GLM程序
3、databeer;inputobsnxyzb;cards;0130102800225112608《金融數(shù)據(jù)挖掘與應用》課程作業(yè)033513330044014400054514410062012270071811210082512280092513300102313290114014410124515420134816425145018450155519470;procglm;modelzb=xy/p;procglm;modelzb=xyx*xx*yy*y/p;procglm;modelzb=xyx*x*x
4、x*x*yx*y*yy*y*y/p;procglm;modelzb=xyx*x*xx*x*yx*y*yy*y*yx*x*x*xx*x*x*yx*x*y*yx*y*y*yy*y*y*y/p;run;2.四種分析模型結果(1)一階趨勢模型DependentVariable:zb源變量自由度平方和均值F值概率值SumofSourceDFSquaresMeanSquareFValuePr>FModel290615.2099345307.60497127.19<.0001Error124274.79007356.2
5、3251CorrectedTotal1494890.00000R-SquareCoeffVarRootMSEzbMean0.9549505.43922818.87412347.000---------------------------------------------------------------------------------------------------------------------------------SourceDFTypeISSMeanSquareFValuePr>F
6、x189541.5655889541.56558251.36<.0001y11073.644351073.644353.010.1081---------------------------------------------------------------------------------------------------------------------------------SourceDFTypeIIISSMeanSquareFValuePr>F8《金融數(shù)據(jù)挖掘與應用》課程作業(yè)x1146
7、52.2435114652.2435141.13<.0001y11073.644351073.644353.010.1081---------------------------------------------------------------------------------------------------------------------------------StandardParameterEstimateErrortValuePr>
8、t
9、Intercept64.0499938033
10、.065399191.940.0766x5.383855650.839475676.41<.0001y6.941998693.998720781.740.1081ObservationObservedPredictedResidual1280.0000000294.9856503-14.98565032260.0000000275.0083707-15.00837073330.0000000342.7309246-12.730