No abstract

We study the problem of detecting and profiling terrorists using a combination of ordinary flat classifiers and relational information. Our starting point is a database for a set of individuals characterized by both "local" attributes such as age and criminal background, and "relational" information such as communications among a subset of the individuals. A subset of the individuals have labels (terrorist or normal people), and we would want to design a classifier that captures the patterns of terrorists and achieves good accuracy in predicting the labels of the remaining part of the database. While "flat" (or attribute-based) classifiers such as decision trees mainly concentrate on using local attributes for object classification, the "guilty by association" simple relational classifier utilizes only connections (relations) among objects for the classification task. We present a hybrid approach that iteratively applies a flat classifier augmented with flattened relational attributes for learning and classification. In particular, we describe our experiments on combining fuzzy C-means clustering with iterative relational classification for terrorist detection from the database. Our hybrid classifier has achieved very good prediction accuracy in the experiments.