• Typo
• Refuse to use central tendency to patch missing values. Instead, assign highest response rate because WOE says so
• Marketing people tell me to force the variable into the model
• Selection bias
• Forgot to segment
• Solely rely on data to segment without consulting the biz side
• Just delete observations with missing values, OK, without studying geometricl boundaries
• Using oversampling, but refuse to weight it back. That boosts lift, right? Let us do 50-50
• Insist random sampling is sufficient, while stratified sampling is critical
• Binning too much, or two little
• Selecting variables without repeated sampling
• Forgot to exclude numeric customer id from the candidate variables. AND,it pops….Well, both Unica and Kxen accepted it, So I see no problem
• When the same variable is sourced by different vendors, did not look up the scales under the same name. Just combine them
• Well, SAS Enterprise Miner gave me this model yesterday
• The binary variable is statistically significant, but there are only 27 event=1, out of ~1mm, since only 27 made some purchases..
• Well, I only have 250 events=1. But I think I can use exact logistic to make it up, all right? I got a PHD in Statistics, Trust me, my professor is OK with it. I just called her.
• Build two-stage model without Heckman adjustment
• Use global mean over the WHOLE customer base to replace missing value on a much smaller universe/subset. So average networth of a high networth client group has 22% worth only 225K
• I just spent the past two days boosting R-square. Now it is 92. Great.
• Forgot to set descending option in proc logistic in SAS
• I think we should hold out missing values when conducting EDA.
• Without proper separation of ‘treatment and control
• Treat business entities and individuals as equal and mix them in the same universe
• Runing clustering without validation
• Running discriminant model without validation. So correct classification rate on development is 89% and that over validation is …35%.(no wonder you finished it in two hours and came here to ask me for a raise)
• Disregard link function in multi-nomil models
• I think this is a better variable: xnew=y*y*y*. It is the top variable dominating others.
• Use standardized coefficient to calculate relative importance, because many people are doing and marketing loves it.
• I tried Goolge Analtyics last Friday. It recommends this variable: click stream density over Thanksgivning weekend, on my web portal, on this item
• Let us treat this matrix as unary so we can apply Euclidean, since that runs faster and has a lot of optimal properties. It makes our life easier
• Let us use score from that model to boost this model and use score from this model to boost it back. Is that what they call neural nets, Jia?