def dim_reduction_mRMR(df, k):
import pymrmr
s = datetime.now()
reduced = pymrmr.mRMR(df, 'MIQ', k)
func.execution_time(s)
# finish Beep sound
func.beeep()
# below is how we can prepare the data to use mRMR
#ordinal_df = pd.DataFrame(pd.np.column_stack([y,X_pad]))
#ordinal_df.rename(columns={0:'class'}, inplace=True)
#for i in range(1,1001):
# ordinal_df.rename(columns={i:'Feat'+str(i)}, inplace=True)
#reduced_ordf = prp.dim_reduction_mRMR(ordinal_df)
return reduced
def lstm_eval(X, y, runtype):
s = datetime.now()
es = EarlyStopping(monitor='val_loss', patience=3, verbose=1)
if (runtype == "10_fold_regular"):
print("Regular 10-fold cross-validation Bi-LSTM")
# 10-fold cross-validation
lstm = neuralNet_10fold(X, y, ep=50, batch=256, netType='lstm', es=es)
print(f"Testing Accuracy: {round(lstm[0],3)}%")
print(f"F-score(macro): {round(lstm[1],3)}")
print(f"F-score(micro): {round(lstm[4],3)}")
print(f"Precision: {round(lstm[5],3)}")
print(f"Recall: {round(lstm[6],3)}")
elif (runtype == "10_fold_grand_mean"):
print("Grand mean 10-fold cross-validation Bi-LSTM")
# grand mean 10-fold cross-validation
acc = 0
fc1 = 0
fc2 = 0
perc = 0
rec = 0
for i in range(0, 30):
print(f"Iteration {i+1}/30")
lstm = neuralNet_10fold(X,
y,
ep=50,
batch=256,
netType='lstm',
es=es)
acc = acc + lstm[0]
fc1 = fc1 + lstm[1]
fc2 = fc2 + lstm[4]
perc = perc + lstm[5]
rec = rec + lstm[6]
print("Grand mean of 10-fold cross-validation BiLSTM (30 runs)")
print(f"Grand mean Accuracy: {round(acc/30,3)}%")
print(f"Grand mean F-score(macro): {round(fc1/30,3)}")
print(f"Grand mean F-score(micro): {round(fc2/30,3)}")
print(f"Grand mean Precision: {round(perc/30,3)}")
print(f"Grand mean Recall: {round(rec/30,3)}")
func.execution_time(s)
def SVM_tuning(X, y):
s = datetime.now()
from sklearn.multiclass import OneVsRestClassifier
clf = svm.SVC(verbose=2)
# Choose some parameter combinations to try
parameters = [{
'kernel': ['rbf'],
'gamma': [1e-2, 1e-3, 1e-4, 1e-5],
'C': [0.001, 0.10, 0.1, 10, 25, 50, 100, 1000]
}, {
'kernel': ['sigmoid'],
'gamma': [1e-2, 1e-3, 1e-4, 1e-5],
'C': [0.001, 0.10, 0.1, 10, 25, 50, 100, 1000]
}, {
'kernel': ['linear'],
'C': [0.001, 0.10, 0.1, 10, 25, 50, 100, 1000]
}]
acc_scorer = make_scorer(f1_score, average='macro')
# acc_scorer = make_scorer(average_precision_score)
from sklearn import preprocessing
# y = preprocessing.label_binarize(y, classes=[1, 2, 3,4,5,6,7,8,9])
#grid = GridSearchCV(estimator=model, param_grid=param_grid, n_jobs=-1)
grid_obj = GridSearchCV(clf,
parameters,
cv=3,
scoring=acc_scorer,
n_jobs=4)
grid_obj = grid_obj.fit(X, y)
# Set the clf to the best combination of parameters
clf = grid_obj.best_estimator_
print(grid_obj.best_params_)
func.execution_time(s)
# finish Beep sound
func.beeep()
return grid_obj.best_params_
def svm_eval(X, y, params, runtype):
s = datetime.now()
svm_params = params
if (runtype == "10_fold_regular"):
print("Regular 10-fold cross-validation SVM")
svm = svm_10fold(X, y, svm_params)
print(f"Testing Accuracy: {round(svm[2],3)}%")
print(f"F-score(macro): {round(svm[3],3)}")
print(f"F-score(micro): {round(svm[6],3)}")
print(f"Precision: {round(svm[7],3)}")
print(f"Recall: {round(svm[8],3)}")
elif (runtype == "10_fold_grand_mean"):
print("Grand mean 10-fold cross-validation SVM")
# grand mean
acc = 0
fc1 = 0
fc2 = 0
perc = 0
rec = 0
for i in range(0, 30):
print(f"Iteration {i+1}/30")
svm = svm_10fold(X, y, svm_params)
acc = acc + svm[2]
fc1 = fc1 + svm[3]
fc2 = fc2 + svm[6]
perc = perc + svm[7]
rec = rec + svm[8]
print("Grand mean of 10-fold cross-validation SVM (30 runs)")
print(f"Grand mean Accuracy: {round(acc/30,3)}%")
print(f"Grand mean F-score(macro): {round(fc1/30,3)}")
print(f"Grand mean F-score(micro): {round(fc2/30,3)}")
print(f"Grand mean Precision: {round(perc/30,3)}")
print(f"Grand mean Recall: {round(rec/30,3)}")
func.execution_time(s)
def dim_reduction_ForwardSelection(X, y, k):
from sklearn import svm
from mlxtend.feature_selection import SequentialFeatureSelector as SFS
from sklearn.metrics import make_scorer
from sklearn.metrics import f1_score
s = datetime.now()
acc_scorer = make_scorer(f1_score, average='macro')
sfs = SFS(svm.SVC(),
k_features=k,
forward=True,
floating=False,
verbose=2,
scoring=acc_scorer,
cv=0)
# Perform SFS
sfs = sfs.fit(X, y)
sfs.k_feature_names_
feat_cols = list(sfs.k_feature_idx_)
print(feat_cols)
func.execution_time(s)
# finish Beep sound
func.beeep()
return feat_cols