An augmentation of random forest through using the principle of justifiable granularity

Le résumé fourni par la source

In recent years, ensemble learning has garnered increasing attention within the field of machine learning. This is primarily because ensemble learning techniques have demonstrated their capability to enhance model performance. Ensemble learning enables the amalgamation of multiple weak learners into a single robust learner, leading to improvements in overall model performance. One of the key benefits of ensemble learning is its ability to reduce model variance, enhance model robustness, and mitigate the risk of overfitting. Additionally, it contributes to reducing data bias and elevating model stability. One well-established ensemble learning algorithm is the Random Forest, which comprises numerous decision trees, typically hundreds of them. Each tree is trained on different subsets of data and a subset of features. This combination enhances the model's robustness and generalization performance. In the context of a regression task with a Random Forest model, results are typically obtained through a simple averaging process. However, this basic averaging approach can introduce significant bias into the results. To address this issue, one can introduce the concept of employing justifiable granularity when handling the prediction results of a Random Forest regression model. This approach offers the advantage of generating interval-valued results instead of numerical ones. The interval values derived through the inclusion of justifiable granularity are calculated based on two fundamental characteristics. Consequently, the interval values possess a certain degree of confidence, and the results in this format offer a level of interpretability and credibility compared to numerical results.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.