KindredSpirit is a one-person project designed to help people connect with others who have similar interests. It uses a transparent, customizable algorithm to match you with people who share your tastes and also suggests content based on these shared preferences. Basically, KindredSpirit helps you meet like-minded people and find new, interesting content.
If you have any questions, comments, concerns or suggestions please feel free to reach out to me at [email protected]
Percentiles are a method for ranking items based on their relative position within a group. For example, being in the 90th percentile means that the item is ranked equal to or higher than 90% of the other items in the group. In the context of KindredSpirit, percentiles are utilized to categorize content based on popularity, rating, and various other metrics. A basic formula for calculating a percentile is as follows:
percentile = round((number_of_items_ranked_at_or_lower / total_number_of_items) * 100)To illustrate, imagine you rated a movie 80 out of 100. If there are 20 movies rated at 80 or lower, and the total number of movies is 50, the percentile would be calculated as:
percentile = round((20 / 50) * 100) = 40For additional information on percentiles, you can visit Wikipedia
Similarity is a metric used to determine how closely two users' ratings align with each other. It is calculated using an inverse transformation of the Euclidean distance between the two users' rating percentiles for each item they have both rated, combined with a custom Jaccard index applied to their sets of ratings. A similarity score of 100% indicates perfect alignment in ratings with at least a quarter of their ratings in common, suggesting the users are ideal kindred spirits. Conversely, a 0% similarity score implies completely opposite tastes. It is important to note that if the users have a very low number of ratings in common, the similarity score will not be displayed or factored into estimated ratings calculations, due to its unreliability.
Wikipedia pages on Euclidean distance and Jaccard index can be found here and here respectively.
The estimated rating is the score you're likely to give to an item, based on your historical ratings and preferences. This is derived from a weighted average of ratings from users with similar tastes, where the weight is determined by how closely their tastes match yours, adjusted by a constant threshold for similarity. It is important to note that if the number of ratings from users with similar tastes is very low, the estimated rating will not be displayed due to its unreliability. A basic formula for an estimated rating percentile is:
estimated_rating_percentile = round(
sum(user_rating_percentile * (user_similarity - similarity_threshold)) / sum((user_similarity - similarity_threshold))
)For example, suppose you come across a movie that you haven't rated. It has been rated by two users whose tastes are similar to yours: the first user's rating is in the 80th percentile with a similarity score of 50% to you, and the second user's rating is in the 55th percentile with a similarity score of 70% to you. If we set the similarity threshold at 40%, the estimated rating is calculated as:
estimated_rating_percentile = round((80 * (50 - 40) + 55 * (70 - 40)) / ((50 - 40) + (70 - 40))) = 61For additional information on weighted average, you can visit Wikipedia
Our search bars support the following syntax options:
plain text: returns search results containing all included words in any order. For example, the query die hard will find both "Die Hard" and "Hard to Die"."quoted text": restricts search results to an exact phrase match. For example,"die hard" will find "Die Hard" but not "Hard to Die".OR: expands search results to include matches for either term. For example,die hard OR die harder will find "Die Hard" as well as "Die Harder".-: excludes results containing a particular word or phrase. For example,die -hard will find "Die Harder" but not "Die Hard".Unfortunately, our search functionality does not currently support typo tolerance or fuzzy searching.
If the estimated rating or similarity score is hidden or represented by a question mark, it indicates that there are insufficient common ratings with other users to provide reliable measures. To enhance accuracy, please consider rating more content. Also, remember that metrics may take a few hours to update following new ratings.
Absolutely! Visit our Tools page for detailed instructions on how to import your ratings.
If you cannot find the content you are looking for, you can copy and paste its respective Wikipedia page link on our Tools page, and it will be automatically loaded into the database.