Construction and evaluation of classifiers for forensic document analysis

TL;DR

This paper develops and evaluates classifiers for forensic document analysis using a statistical approach, introducing a Bayesian cross-validation method to assess classifier performance with limited samples per writer.

Contribution

It presents a novel Bayesian cross-validation technique for evaluating classifiers in forensic document analysis with small sample sizes.

Findings

Classifiers achieved near perfect accuracy in predicting writers.

The Bayesian cross-validation method effectively assesses classifier performance.

The approach is suitable for datasets with many writers and few samples per writer.

Abstract

In this study we illustrate a statistical approach to questioned document examination. Specifically, we consider the construction of three classifiers that predict the writer of a sample document based on categorical data. To evaluate these classifiers, we use a data set with a large number of writers and a small number of writing samples per writer. Since the resulting classifiers were found to have near perfect accuracy using leave-one-out cross-validation, we propose a novel Bayesian-based cross-validation method for evaluating the classifiers.