Acquire Valued Shoppers WEEK 5 Discretization Discretization by
Acquire Valued Shoppers WEEK 5
Discretization § Discretization by binning § Fields § purchaseamount (how much in $ was spent on a transaction) § from ~58000 distinct values into 235 bins § purchasequantity (how many items included in transaction) § from ~2446 distinct values to 455 bins
Sampling § sample from transactions table of those customers that we have train information for § e. g. those having been offered some offer § target 5% of current transaction size § Unable to complete – running rather simple query with big IO since Saturday now § BIG IO PROBLEM ON DATAMINER!
Correlations § transaction date with category and company § Offer. Value (how much saved) with Offer. Acceptance § Offer. Date/Weekday with Offer. Acceptance § Transaction history with Offer. Acceptance § Total Money Spent § Total number of items bought § Number/Percentage of Items bought from the same category/brand/company § ALSO THIS IS STILL RUNNING – IO
http: //i 12 r-studfilesrv. informatik. tumuenchen. de/dmlab 2014/images/thumb/1/1 a /Purchase. Quantity. By. Number. Of. Transactions. p ng/800 px. Purchase. Quantity. By. Number. Of. Transactions. pn g
http: //i 12 r-studfilesrv. informatik. tumuenchen. de/dmlab 2014/images/thumb/3/35 /Acceptance. Rate. By. Date. png/800 px. Acceptance. Rate. By. Date. png
Plans for next week § Brainstorming on feature extraction One row of features per customer learned from history § Any suggestions welcome?
- Slides: 10