def fit(self, X, Y):
self.knnList = []
for column in range(len(Y[0])):
yvar = np.zeros((len(Y), 1))
for row in range(len(Y)):
yvar[row][0] = Y[row][column]
self.knnList.append(KNN(k=10, max_window_size=100, leaf_size=30))
self.knnList[column].fit(X, yvar)
return self
def partial_fit(self, X, y, classes=None):
if len(self.knnList) == 0:
self.knnList = []
for i in range(len(y[0])):
self.knnList.append(KNN(k=10, max_window_size=100, leaf_size=30))
self.L = len(y[0])
for column in range(len(y[0])):
yvar = y[:, column]
self.knnList[column].partial_fit(X, yvar)
return self
from skmultiflow.evaluation.evaluate_prequential import EvaluatePrequential
from skmultiflow.options.file_option import FileOption
from my_classifier import BatchClassifier
dataset = "elec"
# 1. Create a stream
opt = FileOption("FILE", "OPT_NAME", "./data/" + dataset + ".csv", "CSV",
False)
stream = FileStream(opt, -1, 1)
# 2. Prepare for use
stream.prepare_for_use()
# 2. Instantiate the HoeffdingTree classifier
h = [
KNN(k=10, max_window_size=100, leaf_size=30),
HoeffdingTree(),
BatchClassifier(window_size=100, max_models=10),
]
# 3. Setup the evaluator
eval = EvaluatePrequential(pretrain_size=1000,
output_file='result_' + dataset + '.csv',
max_instances=10000,
batch_size=1,
n_wait=500,
max_time=1000000000,
task_type='classification',
show_plot=True,
plot_options=['performance'])
# 4. Run
eval.eval(stream=stream, classifier=h)