A New Perspective on Subset Partitioning in Ensemble Learning

Artificial intelligence technology is indispensable to today's intelligent systems. Ensemble learning, one of the machine learning techniques that is a sub-branch of AI, was studied in this experimental research. Ensemble Learning is created to overcome the individual limitations of existing models and to combine powers for a more optimized final model by working all together. The study is focused on improving the existing Ensemble Learning approach. The innovative aspect of the study is the introduction of a vertical (feature-based) subsetting technique instead of the horizontal (sample-based) subsetting technique used in the process of preparing the subsets for training different weak learners using homogeneous or heterogeneous learning algorithms. Learning from different features allows each model to learn different patterns from one another, therefore, enhancing the ensemble model's evaluation abilities. The proposed approach has been tested on a process that utilizes a combination of a preprocessing step that uses the vertical subsetting method and a training step that brings out the ensemble model using different (heterogeneous) learning algorithms. The obtained results of the experimental study have shown positive effects in terms of F<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf>-Score and can contribute to the literature. Ensemble learning, particularly with the introduction of novel techniques like vertical subsetting, opens the door for new experimental studies that may be initiated for the development of intelligent systems across multiple fields such as financial forecasting, weather forecasting, medical diagnosis, marketing, e-commerce, cybersecurity, natural language processing, image recognition, and autonomous vehicles.