Page 27 - 2024

Ph.D.
(Engineering & Technology)
DEVELOPMENT OF EFFECTIVE PRIVACY PRESERVING
DATA STREAM MINING TECHNIQUE

Ph.D. Scholar : Patel Aniket Rajendrakumar
Research Supervisor : Dr. Kiran R. Amin

Regi. No.: 15146051005
Abstract :
In today’s technology-driven world, online services are extensively used across various
do- mains like social networks, shopping websites, medical services, banking
transactions, and more. These services generate large amounts of data, which are then
analyzed using data mining techniques. Data mining involves extracting valuable
information from massive datasets using methods from machine learning, statistics, and
database management. It goes beyond just examining data; it discovers unexpected
connections and summarizes data creatively for meaningful insights. The results
obtained from data mining are referred to as models or patterns, which can include rules,
graphs, clusters, tree structures, and recurring patterns.

As data continues to grow due to advancements in data acquisition and storage
technolo- gies, data streams have emerged as a significant challenge. Data streams refer
to continuous and dynamic flows of data, like satellite remote sensing systems
generating constant data. Mining patterns and models from such streams is termed data
stream mining. This poses challenges due to the rapid changes and variability in
streaming data. Data stream mining is applied in various fields like mining, storage,
querying, and computation.

Key challenges in data stream mining include developing quick algorithms for streaming
data, identifying rapidly changing patterns, and managing data due to constraints in time
and space. Traditional data mining methods are static and rely on complete data
availability beforehand. With the shift towards data streams, research has transitioned to
data stream mining due to their unique characteristics.

The main issues in data stream mining involve the need for fast mining algorithms for
continuously evolving data, handling fast-changing concepts, and managing the
variability of streaming data. Solutions involve the development of new techniques that
adapt to streaming data, ensuring privacy while mining, and maintaining accuracy in
results.

In response to these challenges, an approach named EPPXGBOOST (Effective Privacy
Preserving eXtreme Gradient Boosting) is proposed. This approach combines Adaptive

01