def plot_comparaison_filter2():
"""
plot a figure comparing the performance using different correction filters
(Raw, Unilateral Limiter, Bilateral Limiter, Time Averaging, Log Sepctrum Averaging)
:return: figure
"""
dir = "log/NY/peps mini"
pattern = r'(internal|access|lock)\\\d{1,2}.csv$'
pattern_valid = r'(3|6|9|12).csv$'
df_report = pd.DataFrame()
filters = [
'Raw', 'ULimiter', 'BLimiter', 'TA:n=5', 'TA:n=10', 'LSA:n=5',
'LSA:n=10'
]
for index, name in enumerate(filters):
print('Filter: ' + name)
utils.construct_set(dir, pattern, pattern_valid, filter=index)
X_train, X_valid, y_train, y_valid = utils.load_train_valid()
cm, report_temp, classes = utils.train(X_train,
X_valid,
y_train,
y_valid,
method='RF',
param={
"max_features": 2,
"n_estimators": 100
})
df_report = df_report.append(report_temp, ignore_index=True)
df_report.index = filters
df_report.plot(kind='bar', rot=0, ylim=(0.8, 0.95))
df_report.to_csv('data/comparaison_filter.csv', sep=';')
def plot_distribution_prob(fig_name):
"""
plot a figure showing the distribution of the max probability
:param fig_name: saved figure name
:return: figure
"""
dir = "log/peps mini"
pattern = r'(internal|access|lock)\\\d{1,2}.csv$'
pattern_valid = r'(3|6|9|12).csv$'
utils.construct_set(dir, pattern, pattern_valid, filter=1)
X_train, X_valid, y_train, y_valid = utils.load_train_valid()
utils.train(X_train,
X_valid,
y_train,
y_valid,
method='RF',
param={
"max_features": 2,
"n_estimators": 100
},
save_prob=True)
utils.plot_max_probablity_distribution('RF')
plt.title(fig_name)
if not os.path.exists(dir_fig):
os.makedirs(dir_fig)
plt.savefig(dir_fig + '/' + fig_name + '.png')
def plot_report(fig_name, plot_cm=False):
"""
plot the comparison result using different ML methods
:param fig_name: saved figure name
:param plot_cm: whether to plot confusion matrix result
:return: figure
"""
dir = "log/peps mini"
pattern = r'(internal|access|lock)\\\d{1,2}.csv$'
pattern_valid = r'(3|6|9|12).csv$'
utils.construct_set(dir, pattern, pattern_valid)
X_train, X_valid, y_train, y_valid = utils.load_train_valid()
methods = ["Logistic", "LDA", "QDA", "KNN", "SVM", "RF", "GBM", "MLP"]
params = [
None, None, None, {
"n_neighbors": 10
}, {
"C": 0.25,
"gamma": 0.5
}, {
"max_features": 2,
"n_estimators": 100
}, {
"n_estimators": 400,
"max_depth": 3
}, {
"hidden_layer_sizes": (16, 8)
}
]
df_report = pd.DataFrame()
for method, param in zip(methods, params):
cm, report_temp, classes = utils.train(X_train,
X_valid,
y_train,
y_valid,
method=method,
param=param)
df_report = df_report.append(report_temp, ignore_index=True)
if plot_cm:
plt.figure()
utils.plot_confusion_matrix(cm, classes, normalize=True)
plt.title(method)
if not os.path.exists(dir_fig + '/methods/'):
os.makedirs(dir_fig + '/methods/')
plt.savefig(dir_fig + '/methods/' + method + '.png')
df_report.set_index('method', inplace=True)
df_report.plot(kind='bar', rot=0, figsize=(16, 6), ylim=(0.6, 1))
plt.title(fig_name)
if not os.path.exists(dir_fig):
os.makedirs(dir_fig)
plt.savefig(dir_fig + '/' + fig_name + '.png')
def deploy_zone_prediction():
"""
generate java model for zone prediction
:return: training result and java model
"""
dir = "log/peps normal"
pattern = r'(left|right|front|back|start|trunk|lock)\\\d{1,2}.csv$'
pattern_valid = r'(3|6|9|12).csv$'
utils.construct_set(dir, pattern, pattern_valid, filter=1)
utils.save_to_csv()
id = 'EightNormal'
dir_path = 'model/'
rf = utils.train_rf(model_id=id, ntrees=25, weight_lock=1)
if not os.path.exists(dir_path):
os.makedirs(dir_path)
rf.download_pojo(path=dir_path, get_genmodel_jar=False)
def plot_TSNE(fig_name):
"""
plot a figure for 2D t-SNE dimension reduction
:param fig_name: saved figure name
:return: figure
"""
dir = "log/peps mini"
pattern = r'(internal|access|lock)\\\d{1,2}.csv$'
pattern_valid = r'(3|6|9|12).csv$'
utils.construct_set(dir, pattern, pattern_valid)
X, y = utils.load_all()
utils.plot_TSNE(X, y)
plt.title(fig_name)
if not os.path.exists(dir_fig):
os.makedirs(dir_fig)
plt.savefig(dir_fig + '/' + fig_name + '.png')
def compare_binary_multi():
"""
plot a figure comparing the binary classification(access, lock) and multiclass classification(left, right, front, back, lock)
:return: figure
"""
dir = "log\\NY\\peps"
pattern = r'(front|left|right|back|lock)\\\d.csv$'
pattern_valid = r'3.csv$'
utils.construct_set(dir, pattern, pattern_valid, filter=1)
cm_list = list()
labels_list = list()
# multi classification
X_train, X_valid, y_train, y_valid = utils.load_train_valid()
cm, report_temp, classes = utils.train(X_train,
X_valid,
y_train,
y_valid,
method='RF',
param={
"max_features": 2,
"n_estimators": 100
})
cm_list.append(cm)
labels_list.append(classes)
# binary classification
y_train = np.array(y_train)
y_train[y_train != 'lock'] = 'access'
y_valid = np.array(y_valid)
y_valid[y_valid != 'lock'] = 'access'
cm, report_temp, classes = utils.train(X_train,
X_valid,
y_train,
y_valid,
method='RF',
param={
"max_features": 2,
"n_estimators": 100
})
cm_list.append(cm)
labels_list.append(classes)
def plot_comparaison_filter(fig_name):
"""
plot a figure illustrating improved performance using RSSI unilateral correction
:param fig_name: saved figure name
:return: figure
"""
dir = "log/peps mini"
pattern = r'(internal|access|lock)\\\d{1,2}.csv$'
pattern_valid = r'(3|6|9|12).csv$'
df_report = pd.DataFrame()
utils.construct_set(dir, pattern, pattern_valid, filter=0)
X_train, X_valid, y_train, y_valid = utils.load_train_valid()
cm, report_temp, classes = utils.train(X_train,
X_valid,
y_train,
y_valid,
method='RF',
param={
"max_features": 2,
"n_estimators": 100
})
df_report = df_report.append(report_temp, ignore_index=True)
utils.construct_set(dir, pattern, pattern_valid, filter=1)
X_train, X_valid, y_train, y_valid = utils.load_train_valid()
cm, report_temp, classes = utils.train(X_train,
X_valid,
y_train,
y_valid,
method='RF',
param={
"max_features": 2,
"n_estimators": 100
})
df_report = df_report.append(report_temp, ignore_index=True)
df_report.index = ['Original', 'Corrected']
df_report.plot(kind='bar', rot=0, ylim=(0.6, 1))
plt.title(fig_name)
if not os.path.exists(dir_fig):
os.makedirs(dir_fig)
plt.savefig(dir_fig + '/' + fig_name + '.png')
import utils
if __name__ == '__main__':
dir = "log/peps normal"
pattern = r'(left|right|front|back|start|trunk|lock)\\\d{1,2}.csv$'
pattern_valid = r'(3|6|9|12).csv$'
utils.construct_set(dir, pattern, pattern_valid, filter=1)
utils.save_to_csv()
X, y = utils.load_all()
X_train, X_valid, y_train, y_valid = utils.load_train_valid()
# compare train result
methods = ["Logistic", "LDA", "QDA", "KNN", "SVM", "RF", "GBM", "MLP"]
params = [
None, None, None, {
"n_neighbors": 10
}, {
"C": 0.25,
"gamma": 0.5
}, {
"max_features": 2,
"n_estimators": 100
}, {
"n_estimators": 400,
"max_depth": 3
}, {
"hidden_layer_sizes": (16, 8)
}
]