Suggested Items with Collaborative Filtering - A List of Recommendations Based on Shared Preferences
In the digital age, finding the perfect event can often feel like searching for a needle in a haystack. But what if there was a tool that could sift through the chaos and present you with a handful of suitable options? That's where collaborative filtering with cosine similarity comes into play.
This innovative method, similar to movie or book recommendation systems, is designed to suggest a few suitable events from a vast database. By reducing the number of suggestions from a gazillion to around 100,000, it helps you navigate the event landscape with ease.
The process begins by creating a table with events as rows and item names as columns. Each cell in the table represents the presence or absence of an event item, with '1' for items present and '0' for items absent.
Next, the cosine similarity between events is calculated using the Cosine Distance or Cosine Similarity method. The cosine similarity score quantifies how closely related two events are, focusing on the orientation of their preference vectors.
The similarity between events is calculated by dividing the dot product of two rows by the product of the magnitudes of both vectors. This results in a similarity score that ranges from -1 (opposite) to 1 (identical), with higher values indicating greater similarity.
Once the similarity scores are calculated, the system identifies events that closely resemble those in your existing list. This process, known as collaborative filtering, helps find events with vectors similar to those in your past list, suggesting relevant new events.
For example, if you've attended a jazz concert in London last week, the system might recommend a blues concert in Manchester this weekend, based on the similarity between the two events.
To ensure the recommendations are not influenced by personal or specific items, only items that occur across at least 20 different lists are included. This way, the suggestions reflect the underlying theme of your list and become more specific once the most common items for a specific type of event are taken into account.
While this feature has been simulated and tested, it has not yet been deployed in a production environment. Nevertheless, the potential for personalized event recommendations is exciting, promising a future where finding the perfect event is just a click away.
[1] Goldberg, D. E., Nichols, J. C., Agrawal, R., & Brujic, V. (1992). Using collaborative filtering for implicit feedback systems. Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval, 187-196.
[2] Resnick, P., Iacovou, N. D., & Lusk, C. (1994). Recommender systems: collaborative filtering of social preferences. Communications of the ACM, 37(11), 76-83.
[3] Sarwar, B., Karypis, G., Konstan, J., & Riedl, J. (2001). Group recommendation algorithms in collaborative filtering. Proceedings of the 24th annual international ACM SIGIR conference on Research and development in information retrieval, 250-257.
[4] Shani, A., & Rafaeli, S. (2005). News recommendation using collaborative filtering. ACM Transactions on the Web, 9(3), 257-278.
Data-and-cloud-computing technology was harnessed in the creation of the event recommendation system, helping it sift through a vast database of events and present the user with a manageable number of suitable options. This innovative technology, based on collaborative filtering with cosine similarity, works similar to movie or book recommendation systems, utilizing cosine similarity to calculate the closeness of events and suggest relevant new events.
