... [ 19 ] to extract customer attributes. Second, to compute the importance of attributes, a neural network classifier is trained and tested. Lastly, a customer network is constructed based on a new similarity function and then partitioned into segments. This new function addresses the limitation...

... network is constructed based on these attributes and predefined networking rules. For networking, a new concept of “topic similarity” is proposed to reflect social meaning in the customer network. Finally, the network is partitioned by modularity clustering, and the resultant clusters are analyzed...