recommend items similar to items that a user has bought, visited, heard, viewed and ranked positively
Demographic filtering
individuals with certain common personal attributes (sex,age, country, etc.) will also have common preferences
Collaborative Filtering
make recommendations to each user based on information provided by those users we consider to have the most in common with them
Hybrid filtering
recommendation methods
description
memory-based
usually use similarity metrics to obtain the distance between two users, or two items, based on each of their ratios
model-based
Use RS information to create a model that generates the recommendations
The main purpose of both memory-based and model-based approaches is to get the most accurate predictions in the tastes of users.
Use dimension-reduction to address data-sparsity problem. e.g. combine LSI and SVD SVD is expensive so it can only be used offline.
Cold Start
Method
New community
encourge users to make ratings
New item
have a set of motivated users who are responsible for rating each new item in the system
New user
The item to item version of the kNN algorithm significantly reduces the scalability problem.
Similarity Measure
KNN
item-to-item, user-to-user
singularity-similarity
singular similarity should be awarded a higher value
RS-tailored SM
superior compared with traditional SM from statistics
JMSD
besides using the numerical information from the ratings (via mean squared differences) also uses the non-numerical information provided by the arrangement of these
Heuristic SM
PIP(Proximity–Impact–Popularity) Fig.1
Evaluation of RS results
description
Quality of the predictions
MAE, RMSE to compare the ratings and predictions
Quality of the set of rcmd
precision, recall F1
Stability
A RS is stable if the predicitions it provides do not change strongly over a short period of time