Practical Predictive Analytics

Poisson (counts)

A Poisson model is used to model counts of things. That could be the number of insurance claims filed in a given month, the number of calls which are received in a call center in a given minute, or the number of orders which are sold for a particular item. The Poisson distribution is the appropriate way for modeling count data since all data is positive and the range of the distribution is bound by 0 and infinity. The classic way of modeling a Poisson model is through the R glm() function using a poisson link function:

model.poisson <- glm(count ~ v1+v2+v3, data=inputdata, family=poisson()) 

Note that the preceding model specified merely shows the model in a generalized form. Do not try to run it since there are no variables ...