數(shù)據(jù)挖掘?qū)嶒?yàn)報(bào)告Weka的數(shù)據(jù)聚類(lèi)分析

甘肅政法學(xué)院本科生實(shí)驗(yàn)報(bào)告〔2〕姓名:學(xué)院:計(jì)算機(jī)科學(xué)學(xué)院專(zhuān)業(yè):信息管理與信息系統(tǒng)班級(jí):實(shí)驗(yàn)課程名稱(chēng):數(shù)據(jù)挖掘?qū)嶒?yàn)日期:指導(dǎo)教師及職稱(chēng):實(shí)驗(yàn)成績(jī):開(kāi)課時(shí)間：2013—2014學(xué)年一學(xué)期甘肅政法學(xué)院實(shí)驗(yàn)管理中心印制實(shí)驗(yàn)題目Weka的數(shù)據(jù)聚類(lèi)分析小組合作姓名班級(jí)學(xué)號(hào)一、實(shí)驗(yàn)?zāi)康?、了解和熟悉K均值聚類(lèi)的步驟2、利用Weka中提供的simpleKmeans方法對(duì)數(shù)據(jù)文件進(jìn)行聚類(lèi)分析，更深刻的理解k均值算法，并通過(guò)對(duì)實(shí)驗(yàn)結(jié)果進(jìn)行觀察分析，找出實(shí)驗(yàn)中所存在的問(wèn)題。二.實(shí)驗(yàn)環(huán)境Win7環(huán)境下的Eclipse三、實(shí)驗(yàn)內(nèi)容在WEKA中實(shí)現(xiàn)K均值的算法，觀察實(shí)驗(yàn)結(jié)果并進(jìn)行分析。四、實(shí)驗(yàn)過(guò)程與分析一實(shí)驗(yàn)過(guò)程 、1、添加數(shù)據(jù)文件翻開(kāi)Weka的Explore，使用Open方拈點(diǎn)擊翻開(kāi)本次實(shí)驗(yàn)所要使用的raff格式數(shù)據(jù)文件“auto93.raff〞2、選擇算法類(lèi)型點(diǎn)擊Cluster中的Choose，選擇本次實(shí)驗(yàn)所要使用的算法類(lèi)型“SimpleKMeans〞**,WekaExplorerIY4蘆。亡白:二|ClufifyCluster加二口TLEit。5**,WekaExplorerIY4蘆。亡白:二|ClufifyCluster加二口TLEit。5白L白白t4itt'ibu.tti：E.^i￡iielLlimClll￡t4t4t'□.osltrsrMg.|Un5t日舊5uk■■*Pah-tbtsifjrBt?Pi1Lk-frdLluaL-im?Miki-I'i-TixalyBu-i-3Cl-aKlBrsE-?OPTICS?SiitpLUN電aiw?Xli?U-iK…■Cabi-Et■■?DBSrari3、得出實(shí)驗(yàn)結(jié)果選中“ClusterMode〞的“Usetrainingset〞，點(diǎn)擊“Start〞按鈕，觀察右邊“Clustereroutput〞給出的聚類(lèi)結(jié)果如下：==Runinformation==Scheme: weka.clusterers.SimpleKMeans-N2-S10Relation: sInstances: 93Attributes:23ManufacturerTypeCity_MPGHighway_MPGAir_Bags_standardDrive_train_typeNumber_of_cylindersEngine_sizeHorsepowerRPMEngine_revolutions_per_mileManual_transmission_availableFuel_tank_capacityPassenger_capacityLengthWheelbaseWidthU-turn_spaceRear_seat_roomLuggage_capacityWeightDomesticclassTestmode:evaluateontrainingdata===Modelandevaluationontrainingset===kMeansNumberofiterations:5Withinclustersumofsquarederrors:Clustercentroids:Mean/Mode:ChevroletMidsize19.073226.317111 5.9024 3.522 173.85374965.85371964.26830 18.6049 5.561193.7805 108.6098 72.3415 41.634129.0202 15.51783517.5611 23.4512StdDevs:N/AN/A 2.39163.0368N/AN/A 1.261 0.901550.3232581.2098370.73N/A 2.49031.0735 11.1232 5.2435 2.9632 1.94622.772 2.4527358.6609N/A10.2372Cluster1Mean/Mode:FordSmall24.961531.269201 4.2301 1.9942120.15385528.84622622.30771 15.1346 4.7115174.8654100.2692 67.0385 36.8462 26.89112.60692722.30770 16.4019StdDevs:N/AN/A 6.07465.7467N/AN/A 0.7301 0.504740.8149484.7019377.1753N/A 3.02040.848 11.2599 5.5735 2.4968 2.3382.7753 2.3975492.4971N/A 7.9863ClusteredInstances41(44%)52(56%)4、修改Seed值5、得出修改Seed值后的實(shí)驗(yàn)結(jié)果===Runinformation=== -N2-S8Relation: sInstances: 93Attributes:23ManufacturerTypeCity_MPGHighway_MPGAir_Bags_standardDrive_train_typeNumber_of_cy「ndersEngine_sizeHorsepowerRPMEngine_revolutions_per_mileManua[transmission_availableFuel_tank_capacityPassenger_capacityLengthWheelbaseWidthU-turn_spaceRear_seat_roomLuggage_capacityWeightDomesticclassTestmode:evaluateontrainingdata===ModelandevaluationontrainingsetkMeansNumberofiterations:6Clustercentroids:Cluster0Mean/Mode:ChevroletMidsize19.073226.317111 5.9024 3.522 173.85374965.85371964.26830 18.6049 5.561193.7805 108.6098 72.3415 41.634129.0202 15.51783517.5611 23.4512StdDevs: N/A N/A2.3916 3.0368N/A N/A1.261 0.9015 50.3232 581.2098370.73N/A 2.4903 1.073511.1232 5.2435 2.9632 1.94622.772 2.4527 358.6609N/A10.2372Cluster1Mean/Mode: Ford Small24.961531.269201 4.2301 1.9942120.15385528.84622622.30771 15.1346 4.7115174.8654 100.2692 67.0385 36.846226.891 12.60692722.30770 16.4019StdDevs: N/A N/A6.0746 5.7467N/A N/A0.7301 0.5047 40.8149 484.7019377.1753N/A 3.0204 0.84811.2599 5.5735 2.4968 2.3382.7753 2.3975 492.4971N/A7.9863ClusteredInstances0 41(44%)1 52(56%)6、對(duì)上述實(shí)驗(yàn)結(jié)果進(jìn)行分析7、顯示散點(diǎn)圖我們?cè)谧笙路健癛esultlist〞列出的結(jié)果上右擊，點(diǎn)“Visualizeclusterassignments〞o彈出的窗口給出了各實(shí)例的散點(diǎn)圖。最上方的兩個(gè)框是選擇橫坐標(biāo)和縱坐標(biāo)，第二行的"color〞是散點(diǎn)圖著色的依據(jù)，默認(rèn)是根據(jù)不同的簇“Cluster〞給實(shí)例標(biāo)上不同的顏色。二、實(shí)驗(yàn)分析本次實(shí)驗(yàn)采用的數(shù)據(jù)文件是"1993NewCarData〞。通過(guò)觀察本次實(shí)驗(yàn)的實(shí)驗(yàn)結(jié)果觀察，我們可以知道：1、Numberofiterations:5,意味著本次實(shí)驗(yàn)把這93條實(shí)例聚成了5類(lèi)，即K=52、這是評(píng)價(jià)聚類(lèi)好壞的標(biāo)準(zhǔn)，數(shù)值越小說(shuō)明同一簇實(shí)例之間的距離越小。如果把“seed〞參數(shù)改一下，得到的這個(gè)數(shù)值就可能會(huì)不一樣。我們應(yīng)該多嘗試幾個(gè)“seed〞，并采納這個(gè)數(shù)值最小的那個(gè)結(jié)果。3、Clustercentroids:"之后列出了各個(gè)簇中心的位置。對(duì)于數(shù)值型的屬性，簇中心就是它的均值〔Mean〕，如加$11「0的數(shù)值型變量Manufacturer的均值19.0732;分類(lèi)型的就是它的眾數(shù)〔Mode〕,如cluster。的分類(lèi)型變量Drive_train_type的眾數(shù)為1,也就是說(shuō)這個(gè)屬性上取值為眾數(shù)值1〔有1個(gè)孩子〕的實(shí)例最多對(duì)于數(shù)值型的屬性還給。 ,出了它在各個(gè)簇里的標(biāo)準(zhǔn)差〔StdDevs〕。最后的〃ClusteredInstances"是各個(gè)簇中實(shí)例的數(shù)目及百分比。五、實(shí)驗(yàn)總結(jié)通過(guò)本次試驗(yàn)，使我對(duì)如