0
$\begingroup$

I want to predict a force plate using plantar pressure. The shape of the force plate data is a 15000x6 array, and the shape of the plantar pressure data is a 15000x89 array. I will use a regression model to predict the force plate data. when collecting data to synchronize the force plate data and plantar pressure, I will do time synchronization between the force plate and plantar pressure app. force plate and plantar pressure data will capture 50 data in 1 seconds.

Force Plate Data:

Data shape : (15000,6)

array([[     4.46733  ,      4.39629  ,    -34.2351   ,  -4077.23     ,
         -6206.81     ,   -874.539    ],
       [     7.65166  ,      2.61174  ,    -49.7356   ,  -4846.76     ,
         -9060.05     ,  -1291.39     ],
       [    11.285    ,     -2.91447  ,    -87.9661   ,  -5412.32     ,
        -16345.2      ,   -213.72     ],
       [    12.7313   ,     -6.48048  ,   -123.094    ,  -5939.48     ,
        -23005.6      ,    443.115    ],
       [    11.6425   ,      0.0259204,   -131.717    ,  -6972.53     ,
        -24651.9      ,  -1112.73     ],
       [    12.3602   ,     10.1988   ,   -139.597    ,  -8544.17     ,
        -26118.8      ,  -3260.79     ],
       [    16.0733   ,     12.1455   ,   -165.01     , -10371.5      ,
        -30873.5      ,  -3643.65     ],
       [    21.1933   ,      8.86926  ,   -210.599    , -12673.2      ,
        -39447.9      ,  -2785.69     ],
       [    24.3619   ,      7.59683  ,   -267.449    , -16170.6      ,
        -50300.9      ,  -2823.35     ]])

enter image description here

Plantar Pressure Data :

Data shape : (15000,89)

array([[0.0, 0.0, 91.0, 100.0, 74.0, 7.0, 0.0, 0.0, 0.0, 0.0, 45.0, 123.0, 126.0, 3.0, 60.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 95.0, 104.0, 75.0, 33.0, 0.0, 0.0, 0.0, 0.0, 57.0, 117.0, 123.0, 113.0, 66.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 58.0, 67.0, 46.0, 16.0, 0.0, 8.0, 0.0, 15.0, 51.0, 122.0, 122.0, 111.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 25.0, 31.0, 64.0, 82.0, 125.0, 124.0, 114.0, 18.0, 0.0, 0.0, 0.0, 24.0, 56.0, 105.0, 116.0, 124.0, 77.0, 0.0, 0.0, 0.0, 71.0, 61.0, 0.0, 0.0, 3.0, 0.0],
[0.0, 0.0, 91.0, 100.0, 74.0, 7.0, 0.0, 0.0, 0.0, 0.0, 45.0, 123.0, 126.0, 3.0, 60.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 95.0, 104.0, 75.0, 33.0, 0.0, 0.0, 0.0, 0.0, 57.0, 117.0, 123.0, 113.0, 66.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 62.0, 68.0, 46.0, 12.0, 0.0, 3.0, 0.0, 15.0, 53.0, 124.0, 125.0, 115.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 25.0, 31.0, 64.0, 82.0, 125.0, 124.0, 114.0, 18.0, 0.0, 0.0, 0.0, 24.0, 56.0, 105.0, 116.0, 124.0, 77.0, 0.0, 0.0, 0.0, 71.0, 61.0, 0.0, 0.0, 3.0, 0.0],
[0.0, 0.0, 91.0, 100.0, 68.0, 6.0, 0.0, 0.0, 0.0, 0.0, 29.0, 118.0, 120.0, 2.0, 50.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 89.0, 100.0, 72.0, 32.0, 0.0, 0.0, 0.0, 0.0, 51.0, 113.0, 118.0, 109.0, 61.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 58.0, 67.0, 46.0, 16.0, 0.0, 8.0, 0.0, 15.0, 51.0, 122.0, 122.0, 111.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 25.0, 31.0, 64.0, 82.0, 125.0, 124.0, 114.0, 18.0, 0.0, 0.0, 0.0, 24.0, 56.0, 105.0, 116.0, 124.0, 77.0, 0.0, 0.0, 0.0, 71.0, 61.0, 0.0, 0.0, 3.0, 0.0]])

Method 1: Input Data : (15000,89) Output Data : (15000,6) enter image description here enter image description here enter image description here

I will get good results on training, but when I try using new data to predict it the results will be bad. then I thought maybe because the plantar pressure data are so similar, it would be hard for the model to recognize it. then I tried to add data sequences to each plantar pressure data. the data sequences that I added are 1-50, the reason I added these data sequences is so that the data on plantar pressure are slightly different from one another so that the model is easier to recognize when I test using new data and I enter 1-50 because every second plantar pressure will capture 50 data. after I added the data sequence to the plantar pressure, the shape of the plantar pressure data is (15000,90)

Plantar Pressure data after adding data sequence: the shape of plantar pressure data before adding data sequence is (15000,89), the shape of plantar pressure data after adding data sequence is (15000,90)

enter image description here

Method 2: Input Data : (15000,90) Output Data : (15000,6)

enter image description here enter image description here enter image description here

so after I added the sequence data to the plantar pressure data, the prediction results from the training data and testing data will be good. my question is whether this method is allowed in data science? If yes, what method I use is called and is there a reference that is the same as the method I used in existing research

Improve this question
$\endgroup$
5
  • $\begingroup$ Sorry for my ignorance. What is the plantar pressure? Is it pressure from walking steps? $\endgroup$
    – Nicolas Martin
    Jan 13 at 14:07
  • $\begingroup$ yes, the data come from walking step $\endgroup$
    – stack offer
    Jan 13 at 14:13
  • $\begingroup$ What is the objective? Is it to predict the behavior of the steps after N hours? It is essential to know because the approach might be very different. $\endgroup$
    – Nicolas Martin
    Jan 13 at 14:59
  • $\begingroup$ yes, to predict behavior of step $\endgroup$
    – stack offer
    Jan 13 at 15:21
  • $\begingroup$ Could you give more details? (use chat if necessary) This is important because steps can vary greatly from person to person. 2D CNN could be better than 1D CNN (pressure captors are in 2D); an interpolation might be enough to explain the steps' behavior, etc. $\endgroup$
    – Nicolas Martin
    Jan 14 at 16:22

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.