Improving F-Score of the imbalance visualized pattern dataset for yield prediction robustness
In a non closed loop manufacturing process, a prediction model of the yield outcome can be achieved by visualizing the temporal historical data pattern generated from the inspection machine, discretize to visualized data patterns, and map them into machine learning algorithm.Our previous study shows...
保存先:
| 主要な著者: | , |
|---|---|
| フォーマット: | Conference or Workshop Item |
| 言語: | English |
| 出版事項: |
2008
|
| 主題: | |
| オンライン･アクセス: | http://repo.uum.edu.my/2854/1/Megat_Norulazmi_Megat_Mohamed_Noor.pdf http://repo.uum.edu.my/2854/ http://www.codata.org/08conf/index.html |
| タグ: |
タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
|
| 要約: | In a non closed loop manufacturing process, a prediction model of the yield outcome can be achieved by visualizing the temporal historical data pattern generated from the inspection machine, discretize to visualized data patterns, and map them into machine learning algorithm.Our previous study shows that combination of under-sampling and over sampling techniques unabel wider range of data sets where SMOTE+VDM and random under-sampling produced robust classifier performance of handling better with different batches of prediction test data.In this paper, the integration of K* entropy base similarity distance function with SMOTE, CNN+Tomek Links and the introduction of SMOTE and SMaRT (Synthetic Majority Replacement Technique)combination, has improved the classifiers F-Score robustness. |
|---|