원문정보
초록
영어
Massive calculation tasks always show as a regular problem in the area of data mining. Many traditional data mining algorithms can only deal with small-scale input data and will run slower or even collapse when the input data increase. The problem above is always a bottleneck of traditional data mining algorithm. Better performance can be achieved if we can transplant these algorithms in cloud computing platform and make them run in parallel. Thus, whether the algorithm can be run in parallel properly or not becomes the key to solve the problem mentioned above. By analyzing the process of local linear regression algorithm, the bottleneck and the aspect which can be parallelized in these algorithms corresponding MapReduced algorithms are proposed, which handle the key problem of efficiency successfully. The research achievements gained in this paper provide a solution for MapReducing algorithms of data mining, and the experiment results show the effectiveness of the solution.
목차
1. Introduction
2. Local Weighted Linear Regression Algorithm
3. Steps Included in the Locally Weighted Linear Regression Algorithm
3.1. Determine Neighboring Data Points
3.2 Local Data Point Weighted Processing
3.3. Determination of Linear Regression Function and Regression Coefficient
3.4. Prediction Calculation
4. Implementation of Local Weighted Linear Regression Algorithm in MapReduce
4.1. Partition of Datanode
4.2. Map Stage
4.3. Reduce Stage
5. Experimental Analysis and Results
6. Conclusion
References