COMBINING VALUE AND PROBABILITY MODELS IN DATABASE MINING

摘要

A technique to produce a marketing campaign is described. The technique scores a data set of prospects using a plurality of models that estimate state transition probabilities for the prospects, with the models based on samples of potential contacts and their responses, and scores the data set with a plurality of valuation models to determine rewards gained from the prospects in the data set. The model combines the probability of the event occurring for the prospects and values of the prospects to provide targetability value estimates for the prospects by using Markov decision processes from outputs of the scoring and solves the Markov decision processes for the prospects.