Scribd Logo
Upload

Welcome to Scribd!

  • 2 views

    Ans in Day 4 Slides

    Uploaded by

    Joel Lim

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd

    Ans in Day 4 Slides

    Uploaded by

    Joel Lim
    2 views5 pages

    Copyright

    © © All Rights Reserved

    Available Formats

    PPTX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    2 views5 pages

    Ans in Day 4 Slides

    Uploaded by

    Joel Lim

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    of 5

    Ans in Ex 4.

    See Excel file bootstrap5ans.xlsx for one possible answer to Q1 to Q3. You
    are encouraged to create your own answers. Due to randomisation, your
    answer will be different.

    3. P(Case i is not selected into a bootstrap sample) =

    4. No. The cases not selected into the bootstrap sample can serve as
    testset when the bootstrap sample is used as the trainset.

    1
    Results of Random Forest on Heart data

    • B = 500, RSF size = 3; OOB overall error = (24 + 30)/297 = 18.18%.

    • Q: How are the confusion matrix results determined?


    • Ans: From OOB data and majority rule.

    • Q: Why did the confusion matrix contain only 297 cases when the dataset
    has 303 cases?
    • Ans: 6 cases has missing values and were omitted as na.action= na.omit

    Source: Chew C.H. (2024) A.I., Analytics & Data Science, Vol. 2. 2
    To check how many times each case is OOB
    among the 500 trees, m.RF.1$oob.times
    P(case i is OOB) = (1 – 1/n)n = (1 – 1/297)297 ≈ 0.367

    Checking:
    Case 1: P(OOB) = 177/500 = 0.354
    Case 2: P(OOB) = 185/500 = 0.37
    Case 3: P(OOB) = 193/500 = 0.386
    …

    Interpretation:
    case 1 is not inside 177 of the 500
    trees in the forest, …

    Source: Chew C.H. (2024) A.I., Analytics & Data Science, Vol. 2. 3
    View RF vote for each case in the dataset via
    m.RF.1$votes
    Q: Consider case 1. Does this mean 56% of
    the 500 trees voted AHD = No and 44% of
    the 500 trees voted AHD = Yes?

    Ans: No. Not 500 trees. Only in those trees


    (approx. 1/3 of 500) for which case 1 is OOB.

    Source: Chew C.H. (2024) A.I., Analytics & Data Science, Vol. 2 & randomForest Documentation. 4
    Caution: m.RF.1$err.rate is not the error rate at each
    tree

    Q: Consider row 4. What is the


    meaning of OOB = 0.27? Does this
    mean the 4th tree OOB error is 27%?

    Ans: OOB error using the first 4 trees


    is 27%

    Source: Chew C.H. (2024) A.I., Analytics & Data Science, Vol. 2 & randomForest Documentation. 5

    You might also like