Market Basket Analysis or Affinity Analysis using Association Rules based model is a cross domain Solution Framework used to discover/identify patterns from transactional data (a master-detail set of line items) and serves many down-stream Business processes like Recommendations, Merchandising/Inventory Planning, Product Assortments etc.
In this solution/revisiting of the MBA process, we decouple the Rule/Pattern identification/discovery phase (finding patterns/rules via Association Rules model build) from the Rule/Pattern KPI calculation phase related to the usefulness evaluation of the patterns (scoring patterns/rules via KPIs).
MBA Rules/Patterns are typically evaluated via the Support, Confidence and Lift KPIs. Experts have advocated for the definition of additional KPIs like Conviction, Imbalance Ratio (IR), Kulc factor (Kulczynski) etc. to identify interesting Rule/Patterns. We define these KPIs as well as many additional custom KPIs which help qualify the Rules.
Shankar Somayajula, Architect - Advanced Analytics, Industry Data Models, Oracle
By registering for this TechCast you give permission for your name and email address to be shared with the presenter and for BIWA User Community so we can inform you of future TechCasts and conferences of interest.