Integrating Socio-Visual Features for Flag Recognition Using Random Forest Algorithms

Abstract: This research presents a machine learning-based system for national flag recognition using the Random Forest (RF) algorithm. The model was trained on a dataset comprising 193 country flags and 41 features, including visual attributes and socio-economic indicators. To enhance model performance, a detailed preprocessing pipeline was applied, removing irrelevant features, converting categorical data, handling missing values, and consolidating sparse classes. The processed data was split into training and testing sets, and a RF classifier was trained using 100 trees. The model achieved a classification accuracy of 94.6%, with a low Out-of-Bag (OOB) error rate of 2.6%. While the system performed well overall, it exhibited sensitivity to class imbalance. This study demonstrates the viability of ensemble learning for symbolic image classification and lays the foundation for more inclusive recognition tools in educational, accessibility, and cross-cultural applications. Keywords: Deep Learning, Random Forest Algorithm, Flag Recognition.