NDCG Evaluation Metric for Recommender Systems – Machine Learning Interviews

Recommender systems are important in sevaral application domains such as e-commerce, finance, healthcare and so on. It is important to come up with evaluation metrics to measure how well a recommender system works. Look at our article on Evaluation metrics for recommender systems for a quick overview. Typical classification and regression metrics measure whether the predicted value is close to the actual value. But they do not account for the order of predictions. To evaluate recommender systems we need to measure how relevant the results are and how good the ordering is. The most popular metric to evaluate a recommender system is the MAP@K metric. This metric tries to measure how many of the recommended results are relevant and are showing at the top. However, the MAP@K metric has some shortcomings. The MAP@K metric focusses on precision – which which of the recommended items are relevant and which are not. It does not take into account how relevant the recommended results are. There

Recommender systems are important in sevaral application domains such as e-commerce, finance, healthcare and so on. It is important to come up with evaluation metrics to measure how well a recommender system works. Look at our article on Evaluation metrics for recommender systems for a quick overview. Typical classification and regression metrics measure whether the predicted value is close to the actual value. But they do not account for the order of predictions. To evaluate recommender systems we need to measure how relevant the results are and how good the ordering is. The most popular metr