How can I perform scikit learn quantile transformation/ normalization on stream object

[medha-chippa]

The following is some reproducible data and code to generate the error:

from river.stream import iter_pandas
from river.metrics import Accuracy
from river import metrics
from river import stream
from river import ensemble
from sklearn.preprocessing import QuantileTransformer

students=pd.DataFrame()

students['r_no']=[1,2,3,4,5]
students['course_no']=['CS_123','CS_456','CS_890','CS_678','CS_123']
students['pass']=[1,1,0,0,1]

x_students=students.drop(['pass'],axis=1)
y_students=students['pass']

c1 = ensemble.AdaptiveRandomForestClassifier(split_confidence=0.01,tie_threshold=0.07, seed=123)
c2 = ensemble.AdaptiveRandomForestClassifier(split_confidence=0.01,tie_threshold=0.07, seed=123)

trans_st1 = QuantileTransformer(n_quantiles=3, random_state=123)
trans_st2 = QuantileTransformer(n_quantiles=3, output_distribution='normal', random_state=456)

strm_st1=iter_pandas(X=x_students,y=y_students)
strm_st2=iter_pandas(X=x_students,y=y_students)

metric1 = metrics.Accuracy()
metric2 = metrics.Accuracy()

y_preds_st1 = []
y_preds_st2 = []

for ((xi1, yi1),(xi2, yi2)) in zip(strm_st1,strm_st2):

new_xi1=trans_st1.fit_transform(xi1)

new_xi2=trans_st2.fit_transform(xi2)

y_pred1 = c1.predict_one(xi1)

y_pred2 = c2.predict_one(xi2)

c1.learn_one(xi1, yi1)

c2.learn_one(xi2, yi2)

metric1.update(y_pred=y_pred1, y_true=yi1)

metric2.update(y_pred=y_pred2, y_true=yi2)

y_preds_st1.append(y_pred1)

y_preds_st2.append(y_pred2)

print(f"Accuracy for c1: {metric1.get()}")

print(f"Accuracy for c2: {metric2.get()}")

How can Quantile transformation be performed on river stream object in an incremental fashion?

[MaxHalford]

You're using a scikit-learn transformer in a River context. Obviously that won't work. I assume you're asking if we could implement a QuantileTransformer in River?

Also, can you please add this code in plaintext instead of screenshots? We can't copy/paste the code as it is. Please use fences to highlight your code.

[medha-chippa]

Thank you so much for responding, The code has been added in the form of Plain Text. Please have a look
However I've not been able to find an equivalent of the QuantileTransformer on river.