Collaborative filtering is a technique used in recommendation systems to analyze user-item interaction data and find similarities between users and items. It does not rely on item attributes or user characteristics, as content-based approaches do. Collaborative filtering is widely employed in various applications, such as suggesting movies, products, or music to users based on their past preferences and behaviour. This paper presents a comprehensive review of collaborative filtering methods for building recommendation systems. Bayesian Factorization Machines (BFM) outperform other methods significantly. Ensembling predictions of models with similar performances proves beneficial. Further, modifications to ALS show promising results. The incorporation of bias parameters in the ALS factorization model improves performance directly. Additionally, we show promising results for the use of deep sigmoid factorization if able to be paired with ALS optimization. Including the item count of the user in the feature vector enhances generalization for BFM