0
$\begingroup$

I'm making a resnet regression model to predict force plate data using smartinsole data. so my data input is smartinsole data and the force plate is the label. below is an example of my data sample.

Insole = np.array([
    [1, 91.523766, 52.848377, 1.6216528, 11.01049, 2.2163177, -5.7556944],
    [2, 91.60506, 52.88966, 1.6523972, 11.210422, 2.1089275, -5.5946937],
    [3, 91.60506, 52.88966, 1.6523972, 11.210422, 2.1089275, -5.5946937],
    [4, 91.60506, 52.88966, 1.6523972, 11.210422, 2.1089275, -5.5946937],
    [5, 92.20641, 53.54877, 1.7195811, 11.204859, 2.732724, -4.8005233],
    [6, 92.48749, 53.836216, 1.6288409, 11.107014, 2.8738887, -5.084718],
    [7, 92.487175, 53.77658, 1.9857383, 10.964879, 2.998679, -5.34794],
    [8, 92.93205, 53.769047, 2.035709, 12.186695, 3.0001752, -5.233251],
    [9, 92.93626, 53.840805, 1.7665762, 12.581867, 3.1149976, -5.373571],
    [10, 92.93626, 53.840805, 1.7665762, 12.581867, 3.1149976, -5.373571]
])

ForceData = np.array([
    [9.27914],
    [9.33246],
    [9.26303],
    [9.30597],
    [9.6594],
    [9.04283],
    [8.88866],
    [8.89956],
    [9.10622],
    [8.99438]
])

I think my data is continuous data, and each data point is similar to each other. Before the training process, I only scaled using the minmaxscaler. My training results have low MSE values and good predictions. but when I try to use data testing, the predictions are less than optimal. I think the problem is with my data, and I don't know which method is suitable for my data type.

Training Prediction Results: enter image description here

Testing Prediction Results: enter image description here

Evaluate metrics scoore: enter image description here

$\endgroup$
2
  • $\begingroup$ Can you provide more details about your model? Like the final layers, and output activation? $\endgroup$ Jun 19, 2023 at 13:42
  • $\begingroup$ I use linear function for last activation layer $\endgroup$ Jun 23, 2023 at 7:02

0

Your Answer

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