def mfcc_feature_raw_data_forest(dsg, iters):
# split into training and test sets
data = dsg.create_X_y_split(genre1=genre1, genre2=genre2)
# do random forest hyperparameter selection -> print best-performing results
best_params, score = random_forest_hyperparameter_selection(data, iters)
results_to_file("MFCC Raw", best_params, score, iters)
def full_feature_raw_data_forest(dsg, iters): ## Explore various results on the FULL set of librosa features. # split into training and test sets data = dsg.create_X_y_split(genre1=genre1, genre2=genre2) # do random forest hyperparameter selection -> print best-performing results best_params, score = random_forest_hyperparameter_selection(data, iters)
def hand_picked_PCA_data_forest(dsg, iters):
# split into training and test sets
data = dsg.create_X_y_split(genre1=genre1, genre2=genre2, usePCA=True)
# do random forest hyperparameter selection -> print best-performing results
best_params, score = random_forest_hyperparameter_selection(data, iters)
results_to_file("Hand Picked PCA", best_params, score, iters)
def hand_picked_info_data_forest(dsg, iters):
# split into training and test sets
X_train, y_train, X_test, y_test = dsg.create_X_y_split(genre1=genre1, genre2=genre2)
data = dsg.create_info_gain_subset(X_train, y_train, X_test, y_test)
# do random forest hyperparameter selection -> print best-performing results
best_params, score = random_forest_hyperparameter_selection(data, iters)
results_to_file("Hand Picked Info", best_params, score, iters)
def chroma_feature_raw_data_forest(iters):
# Load full data set (librosa features)
dsg = DataSetGenerator(subset="small", genre1=genre1, genre2=genre2, libFeatureSets=['chroma_cens', 'chroma_cqt', 'chroma_stft'])
# split into training and test sets
data = dsg.create_X_y_split(genre1=genre1, genre2=genre2)
# do random forest hyperparameter selection -> print best-performing results
best_params, score = random_forest_hyperparameter_selection(data, iters)
results_to_file("Chroma Raw", best_params, score, iters)
def mfcc_feature_best_info_gain_forest(dsg,iters):
# split into training and test sets
X_train, y_train, X_test, y_test = dsg.create_X_y_split(genre1=genre1, genre2=genre2)
# create dataset that uses only most info-gaining features
data = dsg.create_info_gain_subset(X_train, y_train, X_test, y_test)
# do random forest hyperparameter selection
best_params, score = random_forest_hyperparameter_selection(data, iters)
results_to_file("MFCC Info Gain", best_params, score, iters)
def full_feature_PCA_data_forest(dsg, iters):
## Explore various results on full set of librosa features, applying PCA
# split into training and test sets
data = dsg.create_X_y_split(genre1=genre1, genre2=genre2, usePCA=True)
# do random forest hyperparameter selection -> print best-performing results
best_params, score = random_forest_hyperparameter_selection(data, iters)
results_to_file("Full Feature PCA", best_params, score, iters)
def full_feature_best_info_gain_forest(dsg, iters, num_feat=None):
## Explore results on full feature set; subsetted by top info gaining features
# split into training and test sets
X_train, y_train, X_test, y_test = dsg.create_X_y_split(genre1=genre1, genre2=genre2)
# create dataset that uses only most info-gaining features
data = dsg.create_info_gain_subset(X_train, y_train, X_test, y_test, num_feat)
# do random forest hyperparameter selection
best_params, score = random_forest_hyperparameter_selection(data, iters)
results_to_file("Full Feature Info Gain", best_params, score, iters)
