On exploring service failures by joint learning in rational databases

An interesting managerial issue in real service applications is to find all the latent factors that may lead to service failures. To this end, in this work, we establish a joint learning framework to explore service failures in rational databases. Specifically, based on the priori classification of failure types, a general clustering method is introduced for fast grouping the data records, which reveals the similar service failures within one group. Then a greedy algorithm is presented to extract interesting rules from the clustered data set. Experimental results on real call center service failure data show that this joint learning method can exact valuable rules efficiently from rational databases.