Big Semiconductor Manufacturing Data Analysis Using Cloud Technique
Autor: | Ming-Chun Tsai, 蔡明純 |
---|---|
Rok vydání: | 2013 |
Druh dokumentu: | 學位論文 ; thesis |
Popis: | 101 In the semiconductor manufacturing industry, one of the key factors to improve the wafer quality is to analyze the existing logs and find out the probable causative parameters affecting the yield of wafers. Due to the huge amount of data and large amounts of parameters that are recorded in logs, it is difficult for the traditional statistical analysis and relational analysis to process such big data to find out the critical parameters affecting the yields. In order to conquer the analysis bottleneck of big data, we take advantage of the high performance computing of MapReduce and design a novel cloud technique with MapReduce named island-based cloud genetic algorithm (ICGA) to mine the critical information. ICGA is integrated of the cloud genetic algorithm and k nearest neighbor (KNN) classifier. In addition, we adopt the concept of statistics to perform the outlier detection to find out the sensitive parameters. Eventually, the critical parameters discovered by ICGA and sensitive parameters detected by the outlier detection are cross verified to obtain the most discriminative parameters. The obtained most discriminative parameters are used to classify the good and the bad wafers. Experimental results show that these parameters can discriminate between good wafers and bad ones with 100% accuracy. In addition, compared with the standalone GA, ICGA speed up the computation by more than 4 times. |
Databáze: | Networked Digital Library of Theses & Dissertations |
Externí odkaz: |