def load_and_run_classification_experiment(
problem_path,
results_path,
cls_name,
dataset,
classifier=None,
resample_id=0,
overwrite=False,
build_train=False,
predefined_resample=False,
):
"""Load a dataset and run a classification experiment.
Method to run a basic experiment and write the results to files called
testFold<resampleID>.csv and, if required, trainFold<resampleID>.csv.
Parameters
----------
problem_path : str
Location of problem files, full path.
results_path : str
Location of where to write results. Any required directories will be created.
cls_name : str
Determines which classifier to use, as defined in set_classifier. This assumes
predict_proba is implemented, to avoid predicting twice. May break some
classifiers though.
dataset : str
Name of problem. Files must be <problem_path>/<dataset>/<dataset>+"_TRAIN.ts",
same for "_TEST".
classifier : BaseClassifier, default=None
Classifier to be used in the experiment, if none is provided one is selected
using cls_name using resample_id as a seed.
resample_id : int, default=0
Seed for resampling. If set to 0, the default train/test split from file is
used. Also used in output file name.
overwrite : bool, default=False
If set to False, this will only build results if there is not a result file
already present. If True, it will overwrite anything already there.
build_train : bool, default=False
Whether to generate train files or not. If true, it performs a 10-fold
cross-validation on the train data and saves. If the classifier can produce its
own estimates, those are used instead.
predefined_resample : bool, default=False
Read a predefined resample from file instead of performing a resample. If True
the file format must include the resample_id at the end of the dataset name i.e.
<problem_path>/<dataset>/<dataset>+<resample_id>+"_TRAIN.ts".
"""
# Check which files exist, if both exist, exit
build_test = True
if not overwrite:
full_path = (
results_path
+ "/"
+ cls_name
+ "/Predictions/"
+ dataset
+ "/testResample"
+ str(resample_id)
+ ".csv"
)
if os.path.exists(full_path):
build_test = False
if build_train:
full_path = (
results_path
+ "/"
+ cls_name
+ "/Predictions/"
+ dataset
+ "/trainResample"
+ str(resample_id)
+ ".csv"
)
if os.path.exists(full_path):
build_train = False
if build_train is False and build_test is False:
return
if predefined_resample:
X_train, y_train = load_ts(
problem_path + dataset + "/" + dataset + str(resample_id) + "_TRAIN.ts"
)
X_test, y_test = load_ts(
problem_path + dataset + "/" + dataset + str(resample_id) + "_TEST.ts"
)
else:
X_train, y_train = load_ts(problem_path + dataset + "/" + dataset + "_TRAIN.ts")
X_test, y_test = load_ts(problem_path + dataset + "/" + dataset + "_TEST.ts")
if resample_id != 0:
X_train, y_train, X_test, y_test = stratified_resample(
X_train, y_train, X_test, y_test, resample_id
)
if classifier is None:
classifier = set_classifier(cls_name, resample_id, build_train)
run_classification_experiment(
X_train,
y_train,
X_test,
y_test,
classifier,
results_path,
cls_name=cls_name,
dataset=dataset,
resample_id=resample_id,
train_file=build_train,
test_file=build_test,
)
def load_and_run_clustering_experiment(
problem_path,
results_path,
cls_name,
dataset,
clusterer=None,
resample_id=0,
overwrite=False,
format=".ts",
train_file=False,
):
"""Run a clustering experiment.
Method to run a basic experiment and write the results to files called
testFold<resampleID>.csv and, if required, trainFold<resampleID>.csv. This
version loads the data from file based on a path. The
clusterer is always trained on the
Parameters
----------
problem_path : str
Location of problem files, full path.
results_path : str
Location of where to write results. Any required directories will be created
cls_name : str
determines which clusterer to use if clusterer is None. In this
case, set_clusterer is called with this cls_name
dataset : str
Name of problem. Files must be <problem_path>/<dataset>/<dataset>+
"_TRAIN"+format, same for "_TEST"
resample_id : int, default = 0
Seed for resampling. If set to 0, the default train/test split from file is
used. Also used in output file name.
overwrite : boolean, default = False
if False, this will only build results if there is not a result file already
present. If True, it will overwrite anything already there.
format: string, default = ".ts"
Valid formats are ".ts", ".arff", ".tsv" and ".long". For more info on
format, see examples/loading_data.ipynb
train_file: boolean, default = False
whether to generate train files or not. If true, it performs a 10xCV on the
train and saves
"""
# Set up the file path in standard format
if not overwrite:
full_path = (
str(results_path)
+ "/"
+ str(cls_name)
+ "/Predictions/"
+ str(dataset)
+ "/testResample"
+ str(resample_id)
+ ".csv"
)
if os.path.exists(full_path):
build_test = False
if train_file:
full_path = (
str(results_path)
+ "/"
+ str(cls_name)
+ "/Predictions/"
+ str(dataset)
+ "/trainResample"
+ str(resample_id)
+ ".csv"
)
if os.path.exists(full_path):
train_file = False
if train_file is False and build_test is False:
return
# currently only works with .ts
trainX, trainY = load_ts(problem_path + dataset + "/" + dataset + "_TRAIN" + format)
testX, testY = load_ts(problem_path + dataset + "/" + dataset + "_TEST" + format)
if resample_id != 0:
trainX, trainY, testX, testY = stratified_resample(
trainX, trainY, testX, testY, resample_id
)
le = preprocessing.LabelEncoder()
le.fit(trainY)
trainY = le.transform(trainY)
testY = le.transform(testY)
if clusterer is None:
clusterer = set_clusterer(cls_name, resample_id)
run_clustering_experiment(
trainX,
clusterer,
trainY=trainY,
testX=testX,
testY=testY,
cls_name=cls_name,
dataset_name=dataset,
results_path=results_path,
)
