| Summary: | 碩士 === 東海大學 === 工業工程與經營資訊學系 === 101 === The manufacture process of beverage packing material is very complicated and the manufacture speed is also fast. It will bring a large number of defective goods and make the cost up if we don’t master the reasons of quality unusual and to reform them.
We try to use data mining to research the attribution of a Taiwan’s beverage packaging material industry quality problem. First, cleaning and coding data, and then we use Two-Steps Cluster and Self-Organizing Maps and compare the correct rate between them. Second, based on the higher correct rate clustering method, we use decision tree to generalize the rules that cause the quality problem. According to these rules, industries can find out the reason of quality problem swiftly.
The research result shows that the correct rate will be higher after clustering and the result of SOM will be better than Two-Steps Cluster in clustering outcome. After that, we also verify that we can obtain the rules to determine the quality problems by using decision tree. Industries can use these methods to renovate rules so that the newest condition of quality can be commanded.
|
|---|
