Model Selection for Poisson Regression via Association Rules Analysis

Abstract

This study integrates association rules analysis, a methodology for selecting potential interactions, with Poisson regression modeling. Though typically ignored in conventional Poisson regression, interactions are very common in practice. However, selecting a Poisson regression model when many main effects and interactions are involved is problematic. In this study, we develop a model selection framework to address this problem. Specifically, we focus on building an optimal Poisson regression model by (1) discretizing the response and quantitative attributes into levels; (2) exploring via association rules analysis combinations of input variables that have a significant impact on response; (3) selecting potential (low-and high-order) interactions; (4) converting these potential interactions into new variables; and (5) selecting variables from all the input variables and the newly created variables (interactions) to build the optimal Poisson regression model. Our model selection procedure is the first approach to enable a global search for potential interactions and the first to establish the optimal combination of main effects and interaction effects in the Poisson regression model. A real-life example is given for illustration. It is shown that the proposed method finds the optimal model including important interactions that cannot be found by other existing methods.

Multinomial Logit Model Building via TreeNet and Association Rules Analysis: An Application via a Thyroid Dataset

“…In step 3, we select the rules to convert into interaction variables for use in step 4. In our previous work [5,7,25], we recommend selecting the rules with the highest confidence from among all the rules obtained in step 2. Between 30 and 50 rules are selected.…”

Pattern Discovery from Big Data of Food Sampling Inspections Based on Extreme Learning Machine

“…Thus, we transform the variable "Manufacturing date" into four variables, namely, "Spring", "Summer", "Autumn" and "Winter", with each having binary values of true or false. The four new variables are called dummy variables, and they are generated in the same manner as described in [21].…”