基于LS-SVM回归和容斥原理的广告复合效果处理方法
|
摘要
针对可获取数据的不同情况,分别建立了基于最小二乘支持向量机(LS-SVM)回归和容斥原理的广告媒体组合整体效果计算模型以对媒体间广告复合效果进行处理。前者的目的在于解决现有方法处理小样本数据时泛化能力不足的问题;后者则首次讨论了复合效果数据可获取情况下媒体间广告复合效果如何处理的问题。通过某金融机构的实际数据对建立的模型分别进行了验证。结果表明,两个模型都能较好地处理媒体间广告的复合效果,从而准确地计算出广告媒体组合的整体效果。和现有方法的比较表明,基于LSSVM回归的模型在泛化能力上有了明显的改善。
A model based on LS-SVM regression and a model based on inclusion-exclusion principle are constructed to calculate the effect of media mix advertising in two cases:(1)available data are advertising effects of each medium;(2)available data are advertising effects of each medium and duplicate effects of them.The model based on inclusion-exclusion principle is about how to deal with the duplicate effect of media mix advertising when the data of duplicate effects of media are available.An experiment based on practical data of a finance institute is given to verify the proposed models.Verification results show that both proposed models can properly treat the duplicate effects in media mix advertising and thus can accurately calculate the total effect of media mix advertising.Furthermore,the model based on LS-SVM regression with better generalization ability than the existing methods on small sample data caused by the timeliness of advertising is also testified by verification results.
A model based on LS-SVM regression and a model based on inclusion-exclusion principle are constructed to calculate the effect of media mix advertising in two cases:(1)available data are advertising effects of each medium;(2)available data are advertising effects of each medium and duplicate effects of them.The model based on inclusion-exclusion principle is about how to deal with the duplicate effect of media mix advertising when the data of duplicate effects of media are available.An experiment based on practical data of a finance institute is given to verify the proposed models.Verification results show that both proposed models can properly treat the duplicate effects in media mix advertising and thus can accurately calculate the total effect of media mix advertising.Furthermore,the model based on LS-SVM regression with better generalization ability than the existing methods on small sample data caused by the timeliness of advertising is also testified by verification results.
引文
[1]钟颖.广告媒体组合影响因素分析及模型研究[D].长沙:中南大学,2003.
[2]V.Kanti Praasd,L.Winston Ring.Measuring sales effects of some marketing mix variables and their interactions[J].Journal of Marketing Research,1976,8:391~396.
[3]Saunders J.The specification of aggregate market models[J].European Journal of Marketing,1987,21(2):1~47.
[4]Gatignon H,Hanssens D M.Modeling marketing interaction with application to salesforce effectiveness[J].Journal of Marketing Research,1987,15:245~257.
[5]Gopalakrishana S,Chatterjee R.A communications response model for a mature industrial product:Application and implications[J].Journal of Marketing Research,1992,14:189~200.
[6]Montgomery D B,Silk A.Estimating dynamic effects of marketing communications expenditures[J].Management Science,1972,18(10):485~501.
[7]Naik P A,Raman K.Understanding the impact of synergy in multimedia communications[J].Journal of Marketing Research,2003,40:375~388.
[8]徐红敏,王海英,梁瑾,黄帅.支持向量机回归算法及其应用[J].北京石油化工学院学报,2010,18(1):62~66.
[9]Peng X J,Wang Y F.A normal least squares support vector machine(NLS-SVM)and its learning algorithm[J].Neurocomputing,2009,72:3734~3741.
[10]Gencoglu M T,Uyar M.Prediction of flashover voltage of insulators using least squares support vector machines[J].Expert System with Applications,2009,36:10789~10798.
[11]贾嵘等.最小二乘支持向量机回归的HHT在水轮发电机组故障诊断中的应用[J].中国电机工程学报,2006,26(22):128~133.
[12]Michaels J J,Gross J L,Grossman J W,Shier D R.Handbook of discrete and combination mathematics[M].Mathematics,CRC Inc.,Boca Raton,2000.
[2]V.Kanti Praasd,L.Winston Ring.Measuring sales effects of some marketing mix variables and their interactions[J].Journal of Marketing Research,1976,8:391~396.
[3]Saunders J.The specification of aggregate market models[J].European Journal of Marketing,1987,21(2):1~47.
[4]Gatignon H,Hanssens D M.Modeling marketing interaction with application to salesforce effectiveness[J].Journal of Marketing Research,1987,15:245~257.
[5]Gopalakrishana S,Chatterjee R.A communications response model for a mature industrial product:Application and implications[J].Journal of Marketing Research,1992,14:189~200.
[6]Montgomery D B,Silk A.Estimating dynamic effects of marketing communications expenditures[J].Management Science,1972,18(10):485~501.
[7]Naik P A,Raman K.Understanding the impact of synergy in multimedia communications[J].Journal of Marketing Research,2003,40:375~388.
[8]徐红敏,王海英,梁瑾,黄帅.支持向量机回归算法及其应用[J].北京石油化工学院学报,2010,18(1):62~66.
[9]Peng X J,Wang Y F.A normal least squares support vector machine(NLS-SVM)and its learning algorithm[J].Neurocomputing,2009,72:3734~3741.
[10]Gencoglu M T,Uyar M.Prediction of flashover voltage of insulators using least squares support vector machines[J].Expert System with Applications,2009,36:10789~10798.
[11]贾嵘等.最小二乘支持向量机回归的HHT在水轮发电机组故障诊断中的应用[J].中国电机工程学报,2006,26(22):128~133.
[12]Michaels J J,Gross J L,Grossman J W,Shier D R.Handbook of discrete and combination mathematics[M].Mathematics,CRC Inc.,Boca Raton,2000.