The Altair Community is migrating to a new platform to provide a better experience for you. The RapidMiner Community will merge with the Altair Community at the same time. In preparation for the migration, both communities are on read-only mode from July 15th - July 24th, 2024. Technical support via cases will continue to work as is. For any urgent requests from Students/Faculty members, please submit the form linked here.
Options

How to find the worst samples in manufacturing factory

dkpengqiuyangdkpengqiuyang Member Posts: 21 Contributor I
edited December 2018 in Help

greeting,

    I try to do abnormal detection in manufacturing factory to reduce return and repair rate, but the alarm rate is 3 times of return rate, I want to pick the worst samples to set the alarm rate = return rate but failed, the vote operator give me too many “NG possibility =1”samples ( I used 5 Classifier and ok/ng training set to find out the ng samples ), I try to solve this problem by increase Classifier and optimize training set, but the affect is not good enought to lower the alarm rate, I also try to do abnormal detection by distance but the result is even worst than vote.

    so how can I improve the accuracy in rapidminer program ?  I want to pick the worst samples with alarm rate = Specified number.

    thanks.

Tagged:

Answers

  • Options
    Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,635 Unicorn

    Did you try adjusting the threshold of your model?  The default is 0.5 but perhaps in your case a lower threshold would be appropriate.  Search the operator list for "threshold" to see the relevant operators and how to use them.

    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
  • Options
    dkpengqiuyangdkpengqiuyang Member Posts: 21 Contributor I

    hi,

    since I use the vote operator (with 5 classifiers) to divide ok and ng examples, the ng output shall be 0, 0.2, 0.4, 0.6, 0.8, 1, even I move the threshold from 0.5 to 0.9,there are still too many ng samples (compare with the return rate), I am trying to optimaze the model and train set, at the same time I want to find a way to set the alarm rate = return rate to locate the worst samples.  can you help me?

    thanks.

Sign In or Register to comment.