|
|
Weighted frequent pattern tree algorithm of
transform data stream with time |
1. Information Department, Affiliated Hospital of Jiangsu University, Zhenjiang, Jiangsu 212001, China; 2. School of Computer Science and Communication Engineering, Jiangsu University, Zhenjiang, Jiangsu 212013, China |
|
|
Abstract The data distribution in data flow was changed with time. To solve the difficulty of traditional mining algorithm for establishing correlation between transaction time characteristics and relationship characteristics, and the high data storage consumption with time and data as basic unit under different dimensions, the concept of time discontinuity was proposed based on FPTree mining algorithm of affairs. The concept of FPTree node weights was introduced, and the time data was dynamically converted. The mining algorithm was designed based on weighted FPTree, and the data mining association rule of flow was mined. The mining experiments were completed on actual dataset. The results show that compared with traditional FPGrowth algorithm, the proposed algorithm can enhance the average recall by 10% and the average precision by 5% when the computational efficiency is reduced by 20%.
|
|
|
|
|
|
|
|