This document summarizes Max Lin's presentation on using recommendation as classification. It discusses using different classification models like baseline, latent factor, LDA topic, and task view models to predict package recommendations for users. It achieved an AUC of 0.9832, placing 2nd in the Kaggle R Recommendation Engine competition. The models were ensembled using logistic regression and evaluated through cross validation. Features like user and package factors were important. Domain knowledge and data cleaning were also important lessons.