Mining for trigger events with survival analysis

This paper discusses a new application of data mining, quantifying the importance of responding to trigger events with reactive contacts. Trigger events happen during a customer’s lifecycle and indicate some change in the relationship with the company. If detected early, the company can respond to the problem and retain the customer; otherwise the customer may switch to another company. It is usually easy to identify many potential trigger events. What is needed is a way of prioritizing which events demand interventions. We conceptualize the trigger event problem and show how survival analysis can be used to quantify the importance of addressing various trigger events. The method is illustrated on four real data sets from different industries and countries.     

Prevalence of Cyclospora species and other enteric pathogens among children less than 5 years of age in Nepal

Stools from 124 Nepalese children aged 6 to 60 months with diarrhea were examined for organisms of the coccidian genus Cyclospora and for other enteric pathogens. Enterotoxigenic Escherichia coli, Giardia Lamblia, Campylobacter species, Cyclospora species, and Cryptosporidium species were the most common pathogens identified. Cyclospora species were detected in none of 74 children < 18 months of age compared with 6 (12%) of 50 children > or = 18 months of age (P = 0.004).     

Quantifying the indirect effects of a marketing contact

The Internet banner advertising has become an important marketing channel in recent years. There is a strong demand in the industry to quantify the expected return from a marketing contact. Contacts have both direct effects, such as a banner ad prompting the viewer to click and buy, and indirect effects, such as building awareness so that future exposures are more likely to result in a sale. While it is usually straightforward to track and model direct effects, models that fail to account for indirect effects produce inaccurate estimates of the contact value and allocate marketing resources inefficiently. Our research is based on a relational database for banner ads from a major American internet service provider, which records the complete information about the exposures, the clicks and the transactions. The indirect effect of marketing contacts is modeled with a decaying function of the time since the prospect gets into the system. Discrete survival analysis is used to estimate the marginal effects of each additional exposure. The approach is illustrated with a pedagogical example based on the estimates from the internet banner ad model.     

Limitations of nonlinear PCA as performed with generic neuralnetworks

Kramer's (1991) nonlinear principal components analysis (NLPCA) neural networks are feedforward autoassociative networks with five layers. The third layer has fewer nodes than the input or output layers. This paper proposes a geometric interpretation for Kramer's method by showing that NLPCA fits a lower-dimensional curve or surface through the training data. The first three layers project observations onto the curve or surface giving scores. The last three layers define the curve or surface. The first three layers are a continuous function, which we show has several implications: NLPCA "projections" are suboptimal producing larger approximation error, NLPCA is unable to model curves and surfaces that intersect themselves, and NLPCA cannot parameterize curves with parameterizations having discontinuous jumps. We establish results on the identification of score values and discuss their implications on interpreting score values. We discuss the relationship between NLPCA and principal curves and surfaces, another nonlinear feature extraction method.     

Optimal selection of media vehicles using customer databases

This paper investigates the problem where an organization must select among multiple media vehicles for a marketing campaign, and determine how many names from each vehicle to impress (the contact depth). The organization can estimate the return from contacting each prospective customer, and this return decreases as depth increases. Different vehicles have different marginal costs per impression, and may have minimum-spend requirements or activation costs to use the vehicle. Decisions are to be made to maximize profit subject to a constraint on the total amount spent. We proposed an optimization model and two computationally efficient methods that often lead to global optimal solutions under practical assumptions. The model is illustrated with two data sets.