The method of presenting data "Faces of Chernov" and their development using asymmetry

    Chernoff Faces is a diagram of the visual presentation of multivariate data in the form of a human face. Each part of the face: nose, eyes, mouth - represents the value of a specific variable assigned to this part (total 18).

    The main idea is that it’s very natural for a person to look at faces, because all people do it every day. Therefore, data analysis turns out to be a kind of "naturalistic". It is easy to make comparisons and it is easy to identify deviations. Even blondes will be able to perform multivariate analysis of a significant amount of data.

    In 1981, Bernhard Flury and Hans Riedwyl improved the concept and added asymmetry to Chernov’s faces. Thus, the number of variables doubled - up to 36.

    So, each face is an array of 18 elements, each of which takes a value from 0 to 1. The value corresponds to the appearance of the corresponding part of the face. The parameters of the studied objects are reduced to these values. Extremes of real data will be accepted as 0 and 1. Everything else - lying in this gap. Based on the resulting array, a face is constructed.

    A description of the parameters of the face and examples of their use in my entry are here

    or under the cat.

    Here are the parameters that are set for the face:

    1. Eye
    size 2. Pupil size
    3. Pupil position
    4. Tilt of the eye
    5. Horizontal position eyes
    6. The vertical position of the eye
    7. The bend of the eyebrow
    8. The density of the eyebrow
    9. The horizontal position of the eyebrow
    10. The vertical position of the eyebrow
    11. The upper border of the hair
    12. The lower border of the hair
    13. Contour
    14. Darkness of the hair
    15. The slope of the shading of the hair
    16. Nose
    17. The size of the mouth
    18. The curvature of the mouth

    The difficulty lies in the correct comparison of the subjects variables with parts of the face. If you make a mistake, important patterns may go unnoticed.

    Flury gives an example of a successful facial analysis. He analyzed 100 real and 100 fake banknotes according to the parameters of the size of borders, indents and diagonals. Here's what happened:

    Fake banknotes clearly stood out in a separate group. Thus, the analysis revealed different groups of objects.

    Asymmetry allows you to consider objects in progress. The second example shows various parameters in the patients to whom the treatment was applied. The left side of the face shows the values ​​of the parameters before, and the right - after treatment.

    See how the state of the parameters has changed. It is easy to understand to whom and how much better it is, without even delving into the essence of the parameters studied.

    The article Graphical Representation of Multivariate Data by Means of Asymmetrical Faces (by Bernard Flury and Hans Riedwyl) can be read on JSTOR

    If you do not have access, I can send it to you in exchange for an interesting link about which I don’t know yet.

    Also popular now: