Laser-printed document classification using random forest and gray prediction models

This paper presents a classification method for laser-printed documents, integrating the random forest algorithm with the gray prediction model to enhance the accuracy and reliability of forensic document examination. The study utilizes 14 laser printers from five different brands as experimental subjects and extracts 14 key feature parameters such as gray mean, contrast, and distribution symmetry using the ImageXpert analysis system. Classification is done by the random forest algorithm, and the gray prediction model is used to enhance accuracy of classification. Finally, experimental results show that the proposed method achieves high precision or accuracy (96.00% for Chinese characters with fewer strokes and 92.86% for punctuation marks [periods]) for the character and punctuation classification. Compared to traditional classification methods, this approach exhibits superior stability and accuracy. The findings highlight the advantages of non-destructive analysis, efficient classification, and robustness, underscoring its potential as a valuable technological tool for forensic document examination in legal contexts.