The field of movie recommendation systems faces significant obstacles in achieving high accuracy and relevance, primarily due to issues such as cluster overlap and data integration. This work introduces a novel strategy that combines the Improved Kernel Self-organising Map (IKSOM) and Silhouette Clustering methods to address these challenges. The IKSOM algorithm enhances the accuracy of movie recommendations by improving the representation of user and item information in a high-dimensional space. The technique of Silhouette Clustering is utilised to improve the definition of cluster boundaries and minimise the occurrence of overlapping among clusters of movies. The methodology involves combining many data sources through data fusion techniques to improve the recommendation process by including contextual and user-specific information. By incorporating these techniques, the proposed strategy not only improves the precision and comprehensiveness of suggestions but also effectively reduces the duplication among groups, leading to a greater variety of relevant and suitable movie selections. The performance assessment of the proposed system was conducted using the MovieLens_25M dataset. The findings demonstrate a significant improvement in the accuracy of recommendations, with the system achieving an accuracy rate of 98.2%, a precision rate of 96.6%, a recall rate of 95.2%, an F1-score of 95.4%, a Root Mean Square Error (RMSE) of 0.462, and a Mean Absolute Error (MAE) of 0.243. These indicators illustrate the efficacy of the integrated strategy in delivering recommendations of superior quality. This research provides a comprehensive analysis of the potential enhancements that can be achieved in movie recommendation systems through the merging of IKSOM and Silhouette Clustering techniques. The findings suggest that combining advanced clustering methods with reliable data fusion techniques can significantly improve recommendation performance and increase user satisfaction. Future study should explore the integration of contextual variables and current information to improve the adaptability and precision of recommendation systems.