Mining frequent itemsets over data stream has been challenging task. The incoming data from various sources like ecommerce website, click streams, text, audio, weather forecasting etc. are massive unbounded and high speed that it is impractical to store all, process and scan complete data at the same time to extract information. While processing memory and time are the main parameters must be minimum consumed. Thus the paper provides different algorithms for mining over static and dynamic data also known as data stream.

A data stream mining is relevant issue in the context of information gain. The data arrived are large in amount limitless and high rate with time impractical to stock, excavating and testimony at identical measure of time to retrieve intelligence. Sliding window model utilized for frequent pattern mining data stream mining emphasis on recent data and minimum space consumed. In past algorithm window measurement change was steady to concept variation when stagnant and gets smaller when the concept variation happens. Renewed frequent patterns are moderately kept in the current concept whereas the stable transaction is moved out of window. Panes steadily combined to window and performing unnecessary mining for frequent itemsets, conduct is diminishing. Based on the sliding window model the new algorithm named KF_FSW (Kalman Filter based Flexible Sliding Window Model) which utilize Kalman filter function for prediction and measurement approach. The prediction and measurement method is done on basis of already existing information as measure. Thus coagulating the error for accurate position of behavior variation in window size fluctuates in streamed database. Test on standard dataset reports that proposed algorithm coagulates less number of windows for mining and even predicting efficiently the number of count for change ratio captured by occurring change variation.

The word BIG DATA is nothing but huge amount of data generated through all the sources including social networking sites like facebook, twitter, Intstagram etc. this data sometimes may be repetitive that is same person can have record in more than one databases, whereas it is belonging to a single person so those records should be merged. Also sometimes a situation may occur where you need entire history of a person in this case record linkage will make it possible. Many Researches has been done for efficiently linking the records as record linking is becoming important day-by-day since it increases the quality of the data. In this research we are going to focus on algorithm for efficiently linking the records and keeping records secure. The software called FEBRL is used for comparing our algorithm efficiency with previously defined algorithms.