def run(n_modes=1,
fault_prop=.5,
pcs=5200,
repetitions=1,
filename='FFT-PCA-ANN',
batchsize=512):
normadf, faultdf = dp.load_df(n_modes, fault_prop)
pre_process_init = time.perf_counter()
X, y = preprocessor_fft_pca.df_fft_pca(normadf, faultdf, pcs)
pre_process_finish = time.perf_counter()
pre_proc_time = pre_process_finish - pre_process_init
# ---------------------------------------------------------------------------------------------------------------------------------
ann_settings.inputsize = pcs
estimator = KerasClassifier(build_fn=ann_settings.bin_baseline_model,
epochs=20,
batch_size=batchsize,
verbose=0)
dp.validation(X,
y,
estimator,
repetitions,
n_modes,
pre_proc_time,
fault_prop,
filename,
pcs=pcs,
batchsize=batchsize)
def run(n_modes=1,
fault_prop=.5,
pcs=52,
repetitions=1,
filename='PCA-ANN',
batchsize=512):
normal_data, fault1_df = dp.load_df(n_modes, fault_prop)
pre_process_init = time.perf_counter()
# -------------loading data-----------
# Applying PCA
X, y = preprocessor_pca.df_pca(normal_data, fault1_df, pcs, dp.colNames)
pre_process_finish = time.perf_counter()
pre_proc_time = pre_process_finish - pre_process_init
# setup classifier
ann_settings.inputsize = pcs
estimator = KerasClassifier(build_fn=ann_settings.bin_baseline_model,
epochs=20,
batch_size=batchsize,
verbose=0)
dp.validation(X,
y,
estimator,
repetitions,
n_modes,
pre_proc_time,
fault_prop,
filename,
pcs=pcs,
batchsize=batchsize)
def run(n_modes=1, fault_prop=.5, repetitions=1, filename='svm_'):
normadf, faultdf = dp.load_df(n_modes, fault_prop)
# Adding dummy data, labels that mark if a given occurrence is normal or a failure
pre_process_init = time.perf_counter()
faultdf['failure'] = 1
normadf['failure'] = 0
# join both data classes
full_df = normadf.append(faultdf, ignore_index=True)
full_df = full_df.sample(frac=1).reset_index(drop=True)
# Specify the data
X = full_df.iloc[:, 0:52].astype(float)
# Specify the target labels and flatten the array
# y = np_utils.to_categorical(full_df.iloc[:, 13:14])
y = full_df['failure']
pre_process_finish = time.perf_counter()
pre_proc_time = pre_process_finish - pre_process_init
print(filename + ' pre-process finished')
#setup classifier
estimator = LinearSVC(dual=False, verbose=True)
dp.validation(X, y, estimator, repetitions, n_modes, pre_proc_time,
fault_prop, filename)
def run(n_modes=1, fault_prop=.5, pcs=5200, repetitions=1, filename='FFT-PCA-KNN', neighbors=5):
normadf, faultdf = dp.load_df(n_modes, fault_prop)
pre_process_init =time.perf_counter()
X, y = preprocessor_fft_pca.df_fft_pca(normadf, faultdf, pcs)
pre_process_finish =time.perf_counter()
pre_proc_time = pre_process_finish - pre_process_init
# ---------------------------------------------------------------------------------------------------------------------------------
estimator = KNeighborsClassifier(n_neighbors=neighbors)
dp.validation(X, y, estimator, repetitions, n_modes, pre_proc_time, fault_prop,filename,pcs=pcs,n_neghbors=neighbors)
def run(n_modes=1, fault_prop=.5, pcs=52, repetitions=1, filename='PCA-KNN', batchsize=32, neighbors=5):
normal_data, fault1_df = dp.load_df(n_modes, fault_prop)
pre_process_init =time.perf_counter()
# -------------loading data-----------
X, y = preprocessor_pca.df_pca(normal_data, fault1_df, pcs, dp.colNames)
pre_process_finish =time.perf_counter()
pre_proc_time = pre_process_finish - pre_process_init
# setup classifier
estimator = KNeighborsClassifier(n_neighbors=neighbors)
dp.validation(X, y, estimator, repetitions, n_modes, pre_proc_time, fault_prop, filename,pcs=pcs,n_neghbors=neighbors)
def run(n_modes=1,
fault_prop=.5,
pcs=52,
repetitions=1,
filename='fft_pca_svm_'):
normal_data, fault1_df = dp.load_df(n_modes, fault_prop)
pre_process_init = time.perf_counter()
# -------------loading data-----------
X, y = preprocessor_pca.df_pca(normal_data, fault1_df, pcs, dp.colNames)
pre_process_finish = time.perf_counter()
pre_proc_time = pre_process_finish - pre_process_init
print(filename + ' pre-process finished')
#setup classifier
estimator = LinearSVC(dual=False, verbose=True)
dp.validation(X, y, estimator, repetitions, n_modes, pre_proc_time,
fault_prop, filename)
def run(n_modes=1,
fault_prop=.5,
pcs=52,
repetitions=1,
filename='ANN',
batchsize=512):
normadf, faultdf = dp.load_df(n_modes, fault_prop)
# Adding dummy data, labels that mark if a given occurrence is normal or a failure
pre_process_init = time.perf_counter()
faultdf['failure'] = 1
normadf['failure'] = 0
# join both data classes
full_df = normadf.append(faultdf, ignore_index=True)
full_df = full_df.sample(frac=1).reset_index(drop=True)
# Specify the data
X = full_df.iloc[:, 0:52].astype(float)
# Specify the target labels and flatten the array
# y = np_utils.to_categorical(full_df.iloc[:, 13:14])
y = full_df['failure']
# capture pre-process time
pre_process_finish = time.perf_counter()
pre_proc_time = pre_process_finish - pre_process_init
ann_settings.inputsize = pcs # set input_size as the number of principle components
estimator = KerasClassifier(build_fn=ann_settings.bin_baseline_model,
epochs=20,
batch_size=batchsize,
verbose=0)
dp.validation(X,
y,
estimator,
repetitions,
n_modes,
pre_proc_time,
fault_prop,
filename,
batchsize=batchsize)