WEKA Cost Benefit Analysis
Posted by Armando Brito Mendes | Filed under SAD - DSS, software, visualização
Análise de custo benefício para avaliação de modelos
The Cost/Benefit analysis component is a new visualization tool that was released in Weka versions 3.6.2 and 3.7.1. The tool is particularly useful for the analysis of predictive analytic outcomes for direct mail campaigns (or any ranking application where costs are involved). It allows the user to explore various cost/benefit tradeoffs by interactively selecting different population sizes from the ranked list of prospects or by varying the threshold on the predicted probability of the positive class.
The Cost/Benefit analysis tool is available from both the Explorer and Knowledge Flow user interfaces. In the figure below, the Knowledge Flow is being used to build a predictive model for a real-world direct mail application. The data is historical campaign data from a mail out to solicit donations to a charitable organization. The data set contains 47,706 records with 476 variables (summary variables for donor lifetime giving history, overlay demographics etc.). The percentage of donors in the data is approximately 5%. A 10-fold cross-validation is used to generate predictions from a naive Bayes classifier, and these are then passed to the Cost/Benefit analysis tool.
Tags: data mining, WEKA
Comments are closed.
