learn_one() and predict_one() do not seem to be working

[medha-chippa]

Im working on training a HoeffdingTreeClassifier using stream learning. Im using the learn_one() and predict_one() for the same. At the end of the training process, the list containing the predictions is empty. The data is in the form of a pandas dataframe(x data), and y is a pandas series. The iter_pandas() was used to convert the data to a form that is compatible with river.

Please help.

Thank you

[raphaelsty]

Hi @medha-chippa, is it possible to share some data to help you? A small sample of x_bse_new and y?

[raphaelsty]

Thank you, I reproduced your script and I didn't see any particular problem, maybe you have a problem when loading the data? I have an accuracy of 82.5% on the sample you shared. Maybe it will work if you copy paste my code?

import pandas as pd
from river import stream
from river import tree
from river import metrics

X = pd.read_csv("hepatitis_scikit_multiflow.csv")
Y = X["Class"]
X = X.drop("Class", axis = 'columns')

model = tree.HoeffdingTreeClassifier()

X_y = stream.iter_pandas(X, Y)

metric = metrics.Accuracy()

y_preds = []

for x, y in X_y:
    
    y_pred = model.predict_one(x)
    
    model.learn_one(x, y)
    
    metric.update(y_pred=y_pred, y_true=y)
    
    y_preds.append(y_pred)

print(f"Accuracy: {metric.get()}")

Accuracy: 0.825

[medha-chippa]

Thank you so much @raphaelsty , i have been using google colab to work with my data. Do you think that must have caused the issue?

[raphaelsty]

Did it work? I don't think so, it should works well on collab :)

[medha-chippa]

It just worked, the problem was with using stream=stream.iter_pandas(x_train, y_train) and then using for x_i, y_i in "stream" to iterate over the stream. "stream" is a keyword, I had used the keyword to initialize my stream.
Issue resolved. Thank you so much

[raphaelsty]

Cool! Happy that it work, have a good day @medha-chippa :)