Scribd Logo
Upload

Welcome to Scribd!

  • 47 views

    Feature Selection: Mean Square Between Groups Mean Square Within Groups

    Uploaded by

    Hsu Let Yee Hnin
    This document discusses feature selection using statistical tests to identify the features that have the strongest relationship to the target variable. It calculates the F-value for different features to measure their predictive power. The features with the highest F-values are the states of DE, KA, and area code, indicating they have the strongest influence on the target variable compared to other features.

    Copyright:

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

    Feature Selection: Mean Square Between Groups Mean Square Within Groups

    Uploaded by

    Hsu Let Yee Hnin
    47 views4 pages
    This document discusses feature selection using statistical tests to identify the features that have the strongest relationship to the target variable. It calculates the F-value for different features to measure their predictive power. The features with the highest F-values are the states of DE, KA, and area code, indicating they have the strongest influence on the target variable compared to other features.

    Original Title

    Feature_Selection

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document discusses feature selection using statistical tests to identify the features that have the strongest relationship to the target variable. It calculates the F-value for different features to measure their predictive power. The features with the highest F-values are the states of DE, KA, and area code, indicating they have the strongest influence on the target variable compared to other features.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    47 views4 pages

    Feature Selection: Mean Square Between Groups Mean Square Within Groups

    Uploaded by

    Hsu Let Yee Hnin
    This document discusses feature selection using statistical tests to identify the features that have the strongest relationship to the target variable. It calculates the F-value for different features to measure their predictive power. The features with the highest F-values are the states of DE, KA, and area code, indicating they have the strongest influence on the target variable compared to other features.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 4

    Feature Selection

    Features with the strongest relationship with the output variable are selected using statistical
    tests.
    mean square between groups
    F_value =
    mean square within groups
    F_value of different features are calculated.
    Group1 Group2 Group3
    2 10 10
    3 8 13
    7 7 14
    2 5 13
    6 10 15
    Sum 20 40 65
    Mean 4 8 13

    X - mean ( X −mean)2
    Group1 -2 4
    -1 1
    3 9
    -2 4
    2 4
    Group2 2 4
    0 0
    -1 1
    -3 9
    2 4
    Group3 -3 9
    0 0
    1 1
    0 0
    2 4

    Sum of square within groups (SSW) = (4+1+9+4+4)+(4+0+1+9+4)+(9+0+1+0+4)=54


    Observation Mean X – mean ( X −mean)2
    2 8.3 -6.3 40.1
    3 -5.3 28.4
    7 -1.3 1.8
    2 -6.3 40.1
    6 -2.3 5.4
    10 1.7 2.8
    8 -0.3 0.1
    7 -1.3 1.8
    5 -3.3 11.1
    10 1.7 2.8
    10 1.7 2.8
    13 4.7 21.8
    14 5.7 32.1
    13 4.7 21.8
    15 6.7 44.4
    sum 257.3

    Sum of square total (SST) = 257.3


    SST = SSW + SSB
    SSB = SST – SSW = 257.3 – 54 = 203.3
    Degree of freedom between groups = dfb = groups -1 = 3-1 = 2
    Degree of freedom within groups = dfw =observation – group = 15-3=12
    SSB 203.3
    Mean square between groups = = = 101.667
    dfb 2
    SSw 54
    Mean square within groups = = = 4.5
    dfw 12
    mean square between groups 101.667
    F_value = = = 22.59
    mean square within groups 4.5

    F_value for Telecom_Dataset

    No Feature_Name F_value
    1 State_DE 11.69798
    2 State_KA 9.304263
    3 area code 7.305324
    4 State_MI 6.923433
    5 State_OK 6.0412
    6 State_OH 6.02171
    7 total night minutes 5.447898
    8 total intl charge 5.40735
    9 State_AK 5.400937
    10 State_IL 4.457331
    11 total eve charge 4.297015
    12 total night calls 4.290958
    13 total intl minutes 3.784666
    14 customer service 3.458785
    calls
    15 State_AL 2.455233
    16 State_ID 2.455233
    17 State_MA 2.455233
    18 State_MD 2.455233
    19 State_NE 2.455233
    20 total day minutes 1.825441
    21 Voice mail plan 1 1.825441
    22 total intl calls 1.623302
    23 total night charge 1.621888
    24 State_FL 1.182149
    25 State_GA 1.182149
    26 State_HI 1.182149
    27 State_IA 1.182149
    28 State_IN 1.182149
    29 State_MO 1.182149
    30 State_MT 1.182149
    31 State_OR 1.182149
    32 State_VT 1.182149
    33 State_WY 1.182149
    34 International plan 0 1.182149
    35 State_AZ 1.182149
    36 State_NH 1.182149
    37 State_SC 1.182149
    38 Sate_VA 1.182149
    39 total eve minutes 0.775825
    40 total day calls 0.733302
    41 State_CO 0.599227
    42 account length 0.096057
    43 total day charge 0.014973
    44 total eve calls 0.014891
    45 number vmail 0.012495
    messages
    46 International Plan 1 0.012495
    47 Voice main Plan 0 0.012495
    48 State_AR -19.6667
    49 State_LA -19.6667
    50 State_NJ -19.6667
    51 State_RI -19.6667
    52 State_TX -19.6667
    53 State_WI -19.6667
    54 State_WV -19.6667
    55 State_NY -19.667

    You might also like