Feature fusion is an important machine learning technique because it gives classifiers more powerful capabilities in capturing important characteristics of data. Ensemble learning is a technique in designing machine learning models, which provides an improvement in prediction performance. Predicting protein-protein interactions (PPIs) is still a challenging task in accurately determining the interaction or non-interaction between proteins. Machine learning-based approaches are being deeply researched for PPI determination to reduce the cost and time of biological methods. In this study, we develop a novel framework for feature fusion learning and protein-protein interaction prediction. We evaluated our method using three benchmark PPI datasets. In addition, we also conduct extensive comparisons with existing PPI prediction models. The results indicate that our proposed method is better than existing robust methods. Data and our source code are provided via https://gitlab.com/nhanth/FFL-PPIP.git .