基于RBF神經(jīng)網(wǎng)絡(luò)的期貨走勢(shì)預(yù)測(cè)模型研究.pdf

1、分類(lèi)號(hào)UDCF855795密級(jí)——單位代碼!Q151基于RBF神經(jīng)網(wǎng)絡(luò)的期貨走勢(shì)預(yù)測(cè)模型研究指導(dǎo)教師陳如亮劉海勃學(xué)位授予單位職稱(chēng)副教授大連海事大學(xué)申清學(xué)位級(jí)別管理學(xué)碩士學(xué)科與專(zhuān)業(yè)管理科學(xué)與__=程論文完成同期2005午12月25R論文答辯H期2005年3月1】日答辯委員會(huì)主席TheForecastModelofFuturesBasedonRBFNeuralNetworkAbstractToforecastthepriceoffuture

2、sisahotfieldoffuturesstudiesInvestorshopethatitcanhelpthemavoidriskandmakemoreprofitfromfuturesInthisthesis，thewriterusesneuralnetworktechnologytoforecastthefutures’price，buildingaforecastingmodelcalledTheForecastModelof

3、FuturesBasedonRBFneuralnetworkwitllthebackgroundofDalianCommodityExchange，hopingtoserveinvestorsbetterTheresearchworksmainlyonthreeaspects：firstlyitcombinestheregressiontreealgorithmandRBFneuralnetwork，whichusestradition

4、alregressiontreealgorithmtoinitializethecentersofRBFneuralnetwork’shiddenneuralcellsAndthenitcomparesit’scapabilitywithothertraditionalRBFneuralnetwork’slearningalgorithmsSecondlyitmakestheforecastingmodeltrueandexplains

5、howtoselectsamples，therulesforforecastinganderrorscontroletcItisreallyanintegratedplatformthatwecanuseittocollectrawdata，generatesamplesandforecastThirdlyitpresentsusexperimentaldataComparingtothefactualdatacollected，itp

6、rovidesustheanalysisoferrorsandmakessurethatthemodelisusableandreasonablefromtheangleofpromotingbusinessFinallyittellsusthattheRBFneuralnetwork’slearningalgorithmbasedonregressiontreeissuperthanothertraditionalsunderacer

7、taincircumstancewhichthenumberofsamplesisnotverybigandthesample’sdimensionalityisnotverybigtooAtthemeantimetheforecastingmodelisreasonableandineffectonpricepredictinginshorttimethatcangivesomeusefuldecisionmakinginformat