|
Document Type |
: |
Thesis |
|
Document Title |
: |
DEVELOPING DATA STREAM MINING MODEL FOR BIG DATA APPLICATIONS تطوير نموذج تنقيب البيانات المتدفقة لتطبيقات البيانات الكبيرة |
|
Subject |
: |
Faculty of Computing and Information Technology |
|
Document Language |
: |
Arabic |
|
Abstract |
: |
In recent decades, advances in hardware technology have enabled us to automatically record transactions and other crucial information of everyday life, at a rapid rate. Applications which monitor and scrutinize information using sensor networks are the major sources recording massive amount of data with high pace, speed and velocity. These processes usually generate huge amounts of online data which grow and multiply at an extra ordinary rate. Such type of online data is referred as “data stream”. Various knowledge analysis techniques and extraction procedures are studied and investigated to solve the real-world problems. The data mining community use various approaches and methods such as clustering and classification to discover diverse solutions. But all elucidations and answers may not be effectively applied on data streams, since they require online mining which is a continuous process executed in a fashion. Thus, the storage, querying and mining of these data streams, with high speed and huge volume is extremely challenging task.
Stream mining is concerned with extracting knowledge structures represented in models and patterns in non-stopping streams of information. In this thesis, two main contributions are made. Firstly, a data stream mining algorithm named Sliding Window Random Decision Tree (SWRT) for classifying the data stream, is developed. The proposed model adopts sliding window method as data estimator. Secondly, a change detector method is added to improve SWRT performance, by removing drift caused by time change.) ASWRT) is developed, where ADWIN change detector has been used. Both models are verified against accuracy and time. The models are implemented using MOA (Massive Online Analysis) framework. The results of both model are analyzed and compared with each other, to evaluate the effect of employing change detector module. The results showed that, SWRT algorithm achieved 85.78% accuracy with big data of 1,800,000 and consumed total time of 15.41s. It is also noticed that the accuracy of ASWRT algorithm is better, as it attained 98. 88% of accuracy when compared with SWRT, and the total time consumed is 18.80s. |
|
Supervisor |
: |
Dr. Manal Abdullah |
|
Thesis Type |
: |
Master Thesis |
|
Publishing Year |
: |
1439 AH
2018 AD |
|
Added Date |
: |
Sunday, May 6, 2018 |
|
Researchers
| ابتسام حامد المالكي | Almalki, Ebtesam Hamed | Researcher | Master | |
|