注册一亩三分地论坛,查看更多干货!
您需要 登录 才可以下载或查看附件。没有帐号?注册账号 
x
HI everyone:
Sorry my PC does not input Chinese so please bear with the English below:-baidu 1point3acres
I have some questions about making forecast/prediction for a Time Series (TotalMarketing Expense).
we have a Time Series (TS) labeled 'TotalMarketing' that we want to do monthly prediction
This monthly TS ‘TotalMarketing’ is calculated by aggregation/sum of the following 5 Marketing activity groups’ monthly expense $ amount.
MarketingAdvertising
MarketingAgentCommission
MarketingFreeofChargeProductsServices
MarketingOtherExpense
MarketingTradeShows
On top of the above 5 marketing groups’ monthly data, we also have other monthly data/TS that might be correlated with or have an impact on the ‘TotalMarketing’ in a correlation matrix.
My questions are:
Should I use one or some of the 5 marketing groups to forecast TotalMarketing?
in other words, if we know y = x1+x2+x3+x4+x5, should we use any (or some) of the x to forecast y in a multivariate linear regression?
What are the fields/TS in the correlation matrix that we should use to forecast TotalMarketing and why? should I use the TS that has high correlation with TotalMarketing? or low correlation?
How many fields/TS we should chose and why? How do I do dimentions reduction to cut down some of the TS?. Χ
What approaches/models should we use for this kind of prediction?
I have tried PCA to reduce dimensions and am thinking about using Multivariate Linear Regression/Vector Auto Regression. Please advise!
thanks.--
. check 1point3acres for more.
|