Please enable JavaScript.
Coggle requires JavaScript to display documents.
Recommendation system (Plan: (classify post similarity by text and topic,…
Recommendation system
Plan:
classify post similarity by text and topic
extend heuristics with department , location, etc
add collaborative filtering if it would be enough of data
recoment most popular for some company/department
DMRS(Data mining recomendation systems)
colaborative filtering (based on history not on content)
cold start problem
takes a lot of time and resource
types
user-user
execution type
Offline Data mining
Online Data mining
Content based (propose similar item)
product-product (by product similarity matrix)
personalization
may be hard to do good classifier
small history
works good if it is easy to determine context and features of similarity
Rits (Recommendation system in-the-small)
Heuristic
using human intuition about the problem
using data mining during the development
using machine learning to lear the heuristic
Recomend most popular items
no personalization
link to kgtutorial.github.io