The effect of context on misclassification costs in e-commerce applications
作者：
Highlights：
• Conditions under which a contextual predictive model outperforms an un-contextual one in terms of misclassification costs.
• If the unit of analysis is a single customer or a micro-segment, including context reduces the total cost.
• If unit of analysis is a large segment or whole market, including context depends on dependent variable and cost of errors.
• The finer the contextual information the stronger the effect of context on total cost.
• The use of contextual predictive models depend on unity of analysis, dependent variable and costs of incorrect predictions.
摘要
•Conditions under which a contextual predictive model outperforms an un-contextual one in terms of misclassification costs.•If the unit of analysis is a single customer or a micro-segment, including context reduces the total cost.•If unit of analysis is a large segment or whole market, including context depends on dependent variable and cost of errors.•The finer the contextual information the stronger the effect of context on total cost.•The use of contextual predictive models depend on unity of analysis, dependent variable and costs of incorrect predictions.
论文关键词：Data mining,Business intelligence,Knowledge management applications,Context,Predictive model
论文评审过程：Available online 30 March 2013.
论文官网地址：https://doi.org/10.1016/j.eswa.2013.03.009
