Developing Incremental Learning Models with Prototypes

Incremental learning aims to develop models capable of continuously acquiring knowledge, consistent with many real-world scenarios where data evolves and becomes available over time, and may also come with new classes. When performing classification tasks in incremental learning, there is typically no need to retrain models with previous training data. Instead, models can be updated with new data as it becomes available. This study introduces a novel approach, Incremental Random Forest with Prototypes (IRFwP), for incremental learning using a modified random forest model. It constructs data prototypes for each class by clustering training data while maintaining attribute averages and boundaries. The experiments demonstrate the effectiveness of the proposed approach in comparison to several state-of-the-art incremental learning models.