作者: Pallabi Parveen , Pratik Desai , Bhavani Thuraisingham , Latifur Khan
DOI: 10.4108/ICST.COLLABORATECOM.2013.254135
关键词:
摘要: Users' repetitive daily or weekly activities may constitute user profiles. For example, a user's frequent command sequences represent normative pattern of that user. To find patterns over dynamic data streams unbounded length is challenging. this, an unsupervised learning approach proposed in our prior work by exploiting compressed/quantized dictionary to model common behavior sequences. This suffers scalability issues. Hence, this paper, we propose and implement MapReduce-based framework construct quantized dictionary. We show effectiveness distributed parallel solution on benchmark dataset.