Recommender systems are an important component of online platforms because they help people find relevant items in large amounts of data. Recommender systems are used in many areas, such as e-commerce, movie and music streaming services. They learn about a user’s preferences based on their past interactions and make personalized suggestions. The effectiveness of recommender systems depends on several factors, including similarity measures, collaborative filtering strategies, and learning models. This paper presents a comparative evaluation of different similarity functions that are applied to compute relationships between users or items, followed by user-based and item-based collaborative filtering to generate rating predictions. Here we are comparing different machine learning models such as linear regression, k-nearest neighbors, and random forest are used for comparative analysis. Experiments are conducted on benchmark MovieLens100K and MovieLens1M datasets. The performance of all approaches is evaluated using rating prediction metrics, including Mean Absolute Error and Root Mean Squared Error, along with recommendation quality metrics such as Precision, Recall, and F1-score. The study provides clear insights into how different similarity measures, collaborative filtering approaches, and machine learning models influence recommendation performance After evaluation the results shows that for MovieLens-1M dataset for the user-Based collaborative filtering performed better for linear regression model using cosine similarity achieving MAE – 0.789, RMSE – 0.988, Precision – 0.011, Recall – 0.895, F1-score: 0.022 compared with other models.