Securely Mining UARSTP with Recommendation System in Document Streams

Textual documents made and appropriated on the Internet are always showing signs of change in different structures. A large portion of existing works are given to subject demonstrating and the development of individual themes, while sequential relations of topics in progressive documents distributed by a particular user are disregarded. In this paper, with a specific end goal to describe and identify customized and abnormal behaviors of Internet users, we propose Sequential Topic Patterns (STPs) and figure the issue of mining User-aware Rare Sequential Topic Patterns (URSTPs) in document streams on the Internet. They are uncommon all in all yet generally visit for particular users, so can be connected in some genuine situations, for example, real-time monitoring on abnormal user behaviors. We display a gathering of algorithms to tackle this inventive mining issue through three stages: preprocessing to separate probabilistic topics and distinguish sessions for various users, producing all the STP applicants with bolster values for every user by example development, and selecting URSTPs by making user aware rarity analysis on inferred STPs. Here, we also focused on improving the security, performance and accuracy of the system.