Recommender System

Rating Metrics: Evaluate how accurate a recommender is at predicting ratings that users gave to items

・Root Mean Square Error (RMSE)

measure of average error in predicted ratings

・R Squared (R2)

how much of the total variation is explained by the model

・Mean Absolute Error (MAE)

uses absolute value instead of squaring and taking the root of the average

・Explained Variance

how much of the variance in the data is explained by the model

Ranking Metrics: Evaluate how relevant recommendations are for users

・Precision

measures the proportion of recommended items that are relevant

・Recall

measures the proportion of relevant items that are recommended

・Normalized Discounted Cumulative Gain (NDCG)

evaluates how well the predicted items for a user are ranked based on relevance

・Mean Average Precision (MAP)

average precision for each user normalized over all users

Classification metrics: Evaluate binary labels

・Arear Under Curver (AUC)

integral area under the receiver operating characteristic curve

・Logistic loss (Logloss)

the negative log-likelihood of the true labels given the predictions of a classifier