Product Affinity Modeling Task

(Redirected from product placement)
Jump to: navigation, search

A Product Affinity Modeling Task is a predictive modeling task to build a product affinity model (of which products or services sell together).



  • Teradata
    • … Product Affinity is one module in Teradata's comprehensive suite of …

      … Product Affinity Analysis provides marketers with the power to analyze and target customers based on product-centric purchase histories


  • (Gabor, 2002) ⇒ Gabor Melli. (2002). “PredictionWorks' Data Mining Glossary.
    • Affinity Modeling: The generation of a model that predicts which products or services sell together.
    • Business Benefit:
      • Decreases lost revenue from items that are unnecessarily placed together on a price-based promotion.
      • Data Sources: Raw transaction data (transaction id, item, price, qty).
      • Measurements: Maximization of an accuracy function (usually made up of confidence and support sub-measures. Often the maximization is limited to the rules that cover a particular product segment, such as the top 20% revenue/profit generators.
      • Techniques: n-way Correlation, Fisher (F) Statistic, Association Rules. In DMM see co-occurrence module.
      • Issues: one-to-one affinities are the most common to be reported. Many-to-one and many-to-many affinities are also available.
      • See Also: Association, Cross Sell Modeling, Data Mining Task, Diapers and Beer, Market Basket Analysis, and Price Elasticity Modeling.
      • References: Barry, M. and Linoff, G. "Data Mining Techniques". (1997). “Chapter 8 - Market Basket Analysis".
    • e.g. products placement layout