Please enable JavaScript.

Coggle requires JavaScript to display documents.

Provost - Chapter 7 (Expected Value (The expected value is then the…

- - - - where Vr is the value we get from a response
      - Vnr is the value we get from no response
      - (1-Pr(x)) because people either respond or they don't.
      - Benefits of Vr & Vnr need to be determined separately.
      - Using example of customers responding to an offer.
      - Since a customer can only purchase the product by responding to the offer the expected benefit of not targeting her conveniently is zero.
  - - - Are estimated from the tallies in the confusion matrix by computing the rates of the errors and correct decisions.
      - Each cell of the confusion matrix contains a count of the number of decisions corresponding to the corresponding combination of predicted, actual .
      - Expected value calculation we reduce these counts to rates or estimated probabilities by dividing each count by the total number of instances
    - - Will generate a cost-benefit matrix with the same dimensions as a confusion matrix.
      - The cost-benefit matrix specifies, for each (predicted,actual) pair the cost or benefit of making such a decision.
      - Probabilities can be estimated from data, the costs and benefits often cannot
      - Depend on external information provided via analysis of the consequences of decisions in the context of the specific business problem.
      - A false positive occurs when we classify a consumer as a likely responder and therefore target her, but she does not respond (Benefit is negative - money spent)
      - A false negative is a consumer who was predicted not to be a likely responder (so it wasn't offered), but would have bought it if offered (Nothing gained, nothing lost)
      - A true positive is a consumer who is offered the product and buys it. (Benefit = profit)
      - A true negative is a consumer who was not offered a deal and who would not have bought it even if it had been offered (Nothing gained nothing lost)
      - Cost & Benefit matrix is multiplied with matrix of probabilities & summed into a final value = total expected profit
      - Class priors - factoring out the probabilities of seeing each class. Factoring these out allows us to separate the influence of class imbalance from the fundamental predictive power of the model.
- - - - number of mistakes made on negative examples
    - - cost of each mistake made on a positive example will be higher.
- - - - a model that only considers a very small amount of feature information
    - - A decision tree with only one internal node, the root node. A tree limited to one internal node simply means that the tree induction
        selects the single most informative feature to make a decision