3
$\begingroup$

I am testing https://www.physionet.org/challenge/2017/sources/ submission.

I like one of the submission code, which use Xgboost to train the classifier. Training data is in .mat file which can be converted to csv format.

In below code, I have pretrained model xgb.bin, using which I can test any input signal. But I Want to train the model using different data and create my own training model.

Here is the code which predicts class name for given input ecg file

def predict(data):

    #data = io.loadmat(path)['val'][0]

    from numpy import genfromtxt
    data = genfromtxt('testdata/val.csv', delimiter=',')


    features_noise = np.zeros((5, ))

    snr, rr_num, var, fr, fr2 = find_noise_features(data)
    features_noise[0] = snr
    features_noise[1] = rr_num
    features_noise[2] = var
    features_noise[3] = fr
    features_noise[4] = fr2
    features = extract_basic_features(data, 30000)
    features = np.hstack((features, features_noise.reshape(1, -1)))

    mean_ = np.array([15.96300284066109753667, 0.00412371298595770857, 38811.34497233365254942328,
                      0.48050717744965593115, 0.14397582347542958736])
    scale_ = np.array([4.22917401559752281770, 0.00093664880988427878, 62350.76443798459513345733,
                       0.15396567666240373873, 0.07085474966801086349])
    features_noise -= mean_
    features_noise /= scale_

    prediction = 0
    if features_noise[0] < -2.9:
        prediction = 3
    if features_noise[2] > 6.0:
        prediction = 3
    if features_noise[3] > 3.0:
        prediction = 3
    if features_noise[4] < -2.0:
        prediction = 3

    bst = xgb.Booster({'nthread': 4})
    bst.load_model("xgb.bin")

    dfeatures = xgb.DMatrix(features)
    prediction_prob = bst.predict(dfeatures)
    prediction = np.argmax(prediction_prob)

    return prediction

def run(data): 
    prediction = predict(data)
    print(prediction)

I want to train this classifer using my own dataset. Above model is trained on https://www.physionet.org/challenge/2017/training2017.zip dataset.

Looking at above code, do you have any clue how can I train the model using Xgboost.

Here is my thought steps

  1. Calculate features for all training signal in zip file
  2. Calculate feature_noise as it is in above code for all signals
  3. Calculate dfeatures features for each signals (Should I calculate it for each signals or for all signal together)
  4. Train dfeatures for all signals using Xgboost (How?)
  5. Store the xgboost model

Is this correct steps to do?

$\endgroup$
2
$\begingroup$

I might be a little bit late to the party, but I did exactly what you need in the follow-up challenge of CinC2017, where our algorithm gained the 2nd best score on the hidden test set. Our code is available at https://github.com/martinkropf/ecg-classification Paper is available at https://iopscience.iop.org/article/10.1088/1361-6579/aae13e

$\endgroup$
1
  • $\begingroup$ Links to external resources are encouraged, but please add context around the link so your fellow users will have some idea what it is and why it’s there. See here. Thank you. $\endgroup$ – oW_ Mar 10 '20 at 19:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.