Data-Driven Item Recommendation Algorithm
(Redirected from Data-Driven Item Recommendations Algorithm)
Jump to navigation
Jump to search
A Data-Driven Item Recommendation Algorithm is an item recommendations algorithm that is a data-driven ranking algorithm .
- Context:
- It can range from (typically) being a Supervised Item Recommendations Algorithm to being an Unsupervised Item Recommendations Algorithm.
- It can be implemented by a Data-Driven Item Recommendations System (that solves a data-driven item recommendations task).
- It can range from being a Personalized Recommendation Algorithm to being an Non-Personalized Recommendation Algorithm.
- It can range from being a Simple Data-Driven Relevance-based Item Recommendation Algorithm to a Complex Data-Driven Relevance-based Item Recommendation Algorithmn (such as a sequence-aware recommendation algorithm).
- It can range from being a Content-based Item Recommendation Algorithm to being an Interaction-based Item Recommendation Algorithm (to being a Content and Interaction-based Item Recommendation Algorithm.
- It can range from being a Real-Time Informed Item Recommendation Algorithm (with realtime adaptation to relevance scoring requests) to being an Lag-Time Item Recommendation Algorithm.
- It can be a Diversity-Enforcing Item Recommendation Algorithm.
- It can be a Session-Aware Item Recommendation Algorithm.
- It can be a Explaining Item Recommendation Algorithm.
- It can range from being a Collaborative Filtering-based Recommendation Algorithm to being a User Item Content-based Recommendation Algorithm.
- It can utilize Long User Histories (user history).
- It can support Cold-Start User and Cold-Start Items.
- It can debias from historical trends to “predict the future”.
- It can support Large Item Catalogs.
- …
- Example(s):
- a Probabilistic Matrix Factorization (PMF)-based, which explains the latent factors of users and items from the perspective of probability based on the rating matrix.
- a Collaborative Filtering-based Recommendation Algorithm, which applies collaborative filtering.
- a Collaborative Deep Learning (CDL)-based, which performs deep learning for the items’ auxiliary information and collaborative filtering for the rating matrix.
- an additional Stacked Denoising Autoencoders(aSDAE), a hybrid model that combines SDAE and MF, and learns the latent factors from both user-item rating matrix and auxiliary information for users and items.
- a Factorization Machines based ...
- …
- Counter-Example(s):
- See: Information Filtering Algorithm.
References
2010
- (Amatriain et al., 2010) ⇒ Xavier Amatriain, Alejandro Jaimes, Nuria Oliver, and Josep Pujol. (2010). “Data Mining Methods for Recommender Systems.” In: Recommender Systems Handbook.
- QUOTE: In this chapter, we give an overview of the main Data Mining techniques used in the context of Recommender Systems.