def select(self):
warnings.filterwarnings("ignore", category=DeprecationWarning)
# Implemented model selection using CV
i = -1
best = None # Best Model
bestCV = float('inf') # Best Model Score
stateLens = range(self.min_n_components, self.max_n_components +
1) # Num. of States to Test HMM
Fold = KFold() # KFold Obj.
Fold.n_splits = 2 if len(
self.sequences
) > 1 else 1 # If enouph samples: Num Splits in the DataSet (2 Train & Test).
# Loop over diff. State Num for HMM with Cross Val (Train & Test)
for trainIdx, testIdx in Fold.split(self.sequences):
i += 1 # Index
train = combine_sequences(trainIdx, self.sequences) # Train Set
test = combine_sequences(testIdx, self.sequences) # Test Set
model = self.baseSetModel(
stateLens[i], train[0],
train[1]) # Model Trained on Training Set
if not model: continue
try:
cv = model.score(test[0], test[1]) # Model Scored on Test Set
except:
continue
# Get Best CV Model: Lowest CV
if cv < bestCV:
bestCV = cv
best = model
# Return Best Model Cross Validation
return best
def _split_speakers(self, dataset):
# Assume index of the form spkr_uttid
indices = dataset.indices
kfold = KFold(n_splits=self.k, shuffle=True, random_state=self.RND)
index_pairs = map(self.split_uttid, indices)
speakers = np.array(list(set(map(itemgetter(0), index_pairs))))
if len(speakers) < self.k:
logger.warn(
f"Set number of fold equal to number of speakers ({len(speakers)})"
)
self.k = len(speakers)
kfold.n_splits = self.k
for split in kfold.split(speakers):
yield tuple(
np.array([
i for i, idx in enumerate(indices)
if self.split_uttid(idx)[0] in speakers[subset]
]) for subset in split)
@author: titas
"""
import pandas as pd
from sklearn.model_selection import KFold
from sklearn import linear_model
from sklearn import metrics
Data = pd.read_csv(
"https://raw.githubusercontent.com/vincentarelbundock/Rdatasets/master/csv/ISLR/Auto.csv"
)
X = Data["horsepower"].values.reshape(-1, 1) #
y = Data["mpg"].values.reshape(-1, 1)
kf = KFold()
kf.n_splits = 10
ytests = []
ypreds = []
for train_index, test_index in kf.split(Data):
X_train, X_test = X[train_index], X[test_index]
y_train, y_test = y[train_index], y[test_index]
model = linear_model.LinearRegression()
model.fit(X=X_train, y=y_train)
y_pred = model.predict(X_test)
ypreds += list(y_pred)
ytests += list(y_test)
ms_error = metrics.mean_squared_error(ytests, ypreds)
print(ms_error)
