Skip to content

Commit

Permalink
Merge branch 'main' of https://github.com/hmgomes/CapyMOA into main
Browse files Browse the repository at this point in the history
  • Loading branch information
@YibinSun
YibinSun committed Apr 19, 2024
2 parents 841c225 + 1efab4e commit c773ea8
Showing 1 changed file with 27 additions and 28 deletions.
55 changes: 27 additions & 28 deletions notebooks/benchmarking.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -11,25 +11,26 @@

# Library imports
from capymoa.evaluation.evaluation import *
from capymoa.stream import stream_from_file

# MOA/Java imports
from moa.streams import ArffFileStream
from moa.classifiers.meta import AdaptiveRandomForest, StreamingRandomPatches
from moa.classifiers.trees import HoeffdingTree, EFDT
from moa.classifiers.lazy import kNN
from moa.classifiers.bayes import NaiveBayes
from capymoa.base import MOAClassifier

MAX_INSTANCES = 100
OUTPUT_FILE_PATH = "experiments/experiments_MOA_ARF_2.csv"
OUTPUT_FILE_RIVER_PATH = "experiments/experiments_RIVER_2.csv"
OUTPUT_FILE_PATH = "./experiments/experiments_MOA_ARF_2.csv"
OUTPUT_FILE_RIVER_PATH = "./experiments/experiments_RIVER_2.csv"

## Datasets paths
arff_RTG_2abrupt_path = "./data/RTG_2abrupt.arff"
csv_RTG_2abrupt_path = "./data/RTG_2abrupt.csv"


## Function to abstract the test and train loop using RIVER
def test_then_train_RIVER(dataset, model, max_instances=1000, sample_frequency=100):
def run_test_then_train_RIVER(dataset, model, max_instances=1000, sample_frequency=100):
# Start measuring time
start_wallclock_time, start_cpu_time = start_time_measuring()

Expand Down Expand Up @@ -71,28 +72,26 @@ def test_then_train_RIVER(dataset, model, max_instances=1000, sample_frequency=1


def run_MOA_experiment(
arff_path, model=AdaptiveRandomForest(), CLI="", output_file_path=OUTPUT_FILE_PATH
arff_path, model, CLI="", output_file_path=OUTPUT_FILE_PATH
):
with open(output_file_path, mode="a", newline="") as file:
writer = csv.writer(file)

model.getOptions().setViaCLIString(CLI)
model.prepareForUse()

data_arff = ArffFileStream(arff_path, -1)
data_arff.prepareForUse()
data_arff = stream_from_file(path_to_csv_or_arff=arff_path)
model = MOAClassifier(moa_learner=model(), CLI=CLI, schema=data_arff.get_schema())

acc, wallclock, cpu_time, df = test_then_train_evaluation(
results = test_then_train_evaluation(
stream=data_arff,
learner=model,
max_instances=MAX_INSTANCES,
sample_frequency=MAX_INSTANCES,
)
print(
f"{arff_path}, {model.getClass().getName()} {CLI}, {acc:.4f}, {wallclock:.4f}, {cpu_time:.4f}"
f"{arff_path}, \"{model.__str__()} {CLI}\", {results['cumulative'].accuracy():.4f}, {results['wallclock']:.4f}, {results['cpu_time']:.4f}"
)
writer.writerow(
[arff_path, model.getClass().getName() + CLI, acc, wallclock, cpu_time]
[arff_path, '"'+model.__str__() + CLI+'"', results['cumulative'].accuracy(), results['wallclock'], results['cpu_time']]
)


Expand All @@ -104,17 +103,17 @@ def run_RIVER_experiment(

data_csv = pd.read_csv(csv_path).to_numpy()

acc, wallclock, cpu_time, df = test_then_train_RIVER(
acc, wallclock, cpu_time, df = run_test_then_train_RIVER(
dataset=data_csv,
model=model,
max_instances=MAX_INSTANCES,
sample_frequency=MAX_INSTANCES,
)
print(
f"{csv_path}, {model.__class__.__name__} {CLI}, {acc:.4f}, {wallclock:.4f}, {cpu_time:.4f}"
f'{csv_path}, "{model.__class__.__name__} {CLI}", {acc:.4f}, {wallclock:.4f}, {cpu_time:.4f}'
)
writer.writerow(
[csv_path, model.__class__.__name__ + CLI, acc, wallclock, cpu_time]
[csv_path, '"'+model.__class__.__name__ + CLI+'"', acc, wallclock, cpu_time]
)


Expand All @@ -127,57 +126,57 @@ def experiments_MOA():
["dataset", "classifier", "accuracy", "wallclock(s)", "cpu_time(s)"]
)

run_MOA_experiment(arff_path=arff_RTG_2abrupt_path, model=NaiveBayes(), CLI="")
run_MOA_experiment(arff_path=arff_RTG_2abrupt_path, model=HoeffdingTree(), CLI="")
run_MOA_experiment(arff_path=arff_RTG_2abrupt_path, model=EFDT(), CLI="")
run_MOA_experiment(arff_path=arff_RTG_2abrupt_path, model=NaiveBayes, CLI="")
run_MOA_experiment(arff_path=arff_RTG_2abrupt_path, model=HoeffdingTree, CLI="")
run_MOA_experiment(arff_path=arff_RTG_2abrupt_path, model=EFDT, CLI="")
run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path, model=kNN(), CLI=" -w 1000 -k 3"
arff_path=arff_RTG_2abrupt_path, model=kNN, CLI=" -w 1000 -k 3"
)

run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path,
model=AdaptiveRandomForest(),
model=AdaptiveRandomForest,
CLI="-s 5 -o (Percentage (M * (m / 100))) -m 60",
)
run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path,
model=AdaptiveRandomForest(),
model=AdaptiveRandomForest,
CLI="-s 10 -o (Percentage (M * (m / 100))) -m 60",
)
run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path,
model=AdaptiveRandomForest(),
model=AdaptiveRandomForest,
CLI="-s 30 -o (Percentage (M * (m / 100))) -m 60",
)
run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path,
model=AdaptiveRandomForest(),
model=AdaptiveRandomForest,
CLI="-s 100 -o (Percentage (M * (m / 100))) -m 60",
)
run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path,
model=AdaptiveRandomForest(),
model=AdaptiveRandomForest,
CLI="-s 100 -j 4 -o (Percentage (M * (m / 100))) -m 60",
)

run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path,
model=StreamingRandomPatches(),
model=StreamingRandomPatches,
CLI="-s 5 -o (Percentage (M * (m / 100))) -m 60",
)
run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path,
model=StreamingRandomPatches(),
model=StreamingRandomPatches,
CLI="-s 10 -o (Percentage (M * (m / 100))) -m 60",
)
run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path,
model=StreamingRandomPatches(),
model=StreamingRandomPatches,
CLI="-s 30 -o (Percentage (M * (m / 100))) -m 60",
)
run_MOA_experiment(
arff_path=arff_RTG_2abrupt_path,
model=StreamingRandomPatches(),
model=StreamingRandomPatches,
CLI="-s 100 -o (Percentage (M * (m / 100))) -m 60",
)

Expand Down

0 comments on commit c773ea8

Please sign in to comment.