ratingSums

ratingSums is a simple numerical aggregate used in rating systems to summarize the total of individual ratings. It is often used as a complement to the average rating, providing a quick sense of overall volume and intensity of ratings for an item or a user.

Definition and scope: For a set of ratings {r1, r2, ..., rn} where each rj lies within a chosen rating scale (for example, 1 to M), the ratingSum is the sum of those values: ratingSum = r1 + r2 + ... + rn. For a specific item i, ratingSum(i) denotes the sum of all ratings given to that item by different users. For a specific user u, ratingSum(u) denotes the sum of all ratings that user has given across items.

Example: If an item receives ratings 4, 5, 3, and 2, the ratingSum is 14, while the corresponding average rating is 3.5. The ratingSum alone does not convey how many ratings were given, which is why it is commonly analyzed alongside the ratingCount or the average.

Applications and considerations: ratingSums are used to rank items by total received rating values, to feature in machine learning models as an indicator of engagement or popularity, and as an input in recommender systems. They can aid in anomaly detection when sums are unusually high or low for the number of ratings observed. Because ratingSum grows with the number of ratings, comparing sums across items with different rating counts requires normalization, such as using the average rating or a weighted sum that incorporates ratingCount.

Variants and extensions: Weighted ratingSum incorporates confidence or voter reliability through weights. Normalized or per-item per-user sums can be used to compare across items with differing rating volumes. See also rating average, rating count, rating variance, and related score aggregations.