Improving Cosine Similarity Performance for Large Datasets Using Optimized Data Structures and Algorithms