Mobasher et al. [53] present a system for web personalization based on association rules mining. Their system identifies association rules from pageviews cooccurrences based on users navigational patterns. Their approach outperforms a
kNN-based recommendation system both in terms of precision and coverage. Smyth
et al. [68] present two different case studies of using association rules for RS. In the
first case they use the a priori algorithm to extract item association rules from user
profiles in order to derive a better item-item similarity measure. In the second case,
they apply association rule mining to a conversational recommender. The goal here
is to find co-occurrent critiques – i.e. user indicating a preference over a particular
feature of the recommended item. Lin et al. [49] present a new association mining
algorithm that adjusts the minimum support of the rules during mining in order to
obtain an appropriate number of significant rule therefore addressing some of the
shortcomings of previous algorithms such as the a priori. They mine both association rules between users and items. The measured accuracy outperforms previously
reported values for correlation-based recommendation and is similar to the more
elaborate approaches such as the combination of SVD and ANN.
Mobasher et al. [53] present a system for web personalization based on association rules mining. Their system identifies association rules from pageviews cooccurrences based on users navigational patterns. Their approach outperforms a
kNN-based recommendation system both in terms of precision and coverage. Smyth
et al. [68] present two different case studies of using association rules for RS. In the
first case they use the a priori algorithm to extract item association rules from user
profiles in order to derive a better item-item similarity measure. In the second case,
they apply association rule mining to a conversational recommender. The goal here
is to find co-occurrent critiques – i.e. user indicating a preference over a particular
feature of the recommended item. Lin et al. [49] present a new association mining
algorithm that adjusts the minimum support of the rules during mining in order to
obtain an appropriate number of significant rule therefore addressing some of the
shortcomings of previous algorithms such as the a priori. They mine both association rules between users and items. The measured accuracy outperforms previously
reported values for correlation-based recommendation and is similar to the more
elaborate approaches such as the combination of SVD and ANN.