Does the existing recommendation rely solely on data that includes 'browsing/purchase records', while missing 'user behaviour intentions' (such as duration of stay, reasons for abandoning items after adding to cart, needs mentioned during consultations), 'contextual information' (such as holidays, regions, device types), and 'user preference tags' (such as style, price sensitivity, material preferences)? Is there a 'cold start' problem (where new users/new products lack data support, leading to random recommendations), and how can this be addressed through 'temporary tag completion' (such as new user surveys, automatic tagging of product attributes)?
The recommendations based on 'browsing/ purchase records' is a traditional recommender systems, solely based on historical data ( purchase records etc). This can often lead to monotonous personalized experienced. Over the years, the systems have developed into serendipitous recommender systems. The serendipitous is a surprise driven recommender systems which has a cold start problem as the system does not have any exisiting data of the user. It can offer recommendations based on the peer/ social cricles as they might have similar interests. New user surveys is something people are not very much interesed to participate with due to time/ collection of personal data. Also, the user surveys that appear in between social media - Youtube/ Instagram have an intrusion effect.
- Badges
- Science topic
- Similar topics
- Computer Communications (Networks)
- Simulators