<p>This work introduces a novel multi-view ensemble learning (MEL) framework that utilizes Louvain-based view construction to tackle the challenges of high-dimensional datasets. By partitioning features into multiple views and leveraging their diversity, the method enhances classification accuracy and robustness. The framework involves computing a correlation matrix, constructing a correlation graph, and applying the Louvain algorithm for feature set partitioning, followed by performance-weighted ensemble learning. Extensive trials on ten high-dimensional benchmark datasets compared the suggested approach with eight advanced feature set partitioning (FSP) techniques using a support vector machine classifier. The Louvain-based MEL framework demonstrated superior accuracy and efficiency, supported by statistical analysis via the Friedman ranking test. The results highlight the framework’s ability to reduce feature redundancy and improve data view diversity, offering valuable insights for optimizing machine learning models through effective feature partitioning and ensemble learning strategies.</p>