def getTopModelFeatures(exp_id, size, train_test, fold_id):
exp = updateCurrentExperiment(exp_id)
directory = exp.output_dir()
if fold_id != 'None' and fold_id != 'all':
directory = path.join(directory, fold_id)
directory = path.join(directory, train_test)
filename = path.join(directory, 'model_coefficients.csv')
with open(filename, 'r') as f:
coefficients_df = pd.read_csv(f,
header=0,
index_col=0,
nrows=int(size))
coefficients = list(coefficients_df['mean'])
features_ids = coefficients_df.index
tooltip_data = []
user_ids = []
for feature_id in features_ids:
query = session.query(FeaturesAlchemy)
query = query.filter(FeaturesAlchemy.id == int(feature_id))
row = query.one()
tooltip_data.append(row.name)
user_ids.append(row.user_id)
barplot = BarPlot(user_ids)
dataset = PlotDataset(coefficients, None)
score = exp.exp_conf.core_conf.classifier_conf.featureImportance()
if score == 'weight':
dataset.set_color(colors_tools.red)
barplot.add_dataset(dataset)
return jsonify(barplot.to_json(tooltip_data=tooltip_data))
def _gen_binary_histogram(self):
self.barplot = BarPlot(['0', '1'])
for label, dataset in self.plot_datasets.items():
if len(dataset.values) > 0:
num_0 = sum(dataset.values == 0)
num_1 = sum(dataset.values == 1)
hist_dataset = PlotDataset([num_0, num_1], label)
hist_dataset.set_color(dataset.color)
self.barplot.add_dataset(hist_dataset)
def _gen_histogram(self):
# 10 equal-width bins computed on all the data
_, bin_edges = np.histogram(self.all_values, bins=10, density=False)
x_labels = ['%.2f - %.2f' % (bin_edges[e], bin_edges[e+1])
for e in range(len(bin_edges) - 1)]
self.barplot = BarPlot(x_labels)
for label, dataset in self.plot_datasets.items():
if len(dataset.values) > 0:
hist, _ = np.histogram(dataset.values, bins=bin_edges,
density=False)
hist_dataset = PlotDataset(hist, label)
hist_dataset.set_color(dataset.color)
self.barplot.add_dataset(hist_dataset)
def getFamiliesBarplot(annotations_id, iteration, label):
iteration = None if iteration == 'None' else int(iteration)
family_counts = annotations_db_tools.getFamiliesCounts(session,
annotations_id,
iter_max=iteration,
label=label)
df = pd.DataFrame({
'families': list(family_counts.keys()),
'counts': [family_counts[k] for k in list(family_counts.keys())]
})
matrix_tools.sort_data_frame(df, 'families', ascending=True, inplace=True)
barplot = BarPlot(list(df['families']))
dataset = PlotDataset(list(df['counts']), 'Num. Instances')
dataset.set_color(colors_tools.get_label_color(label))
barplot.add_dataset(dataset)
return jsonify(barplot.to_json())
def getClusterStats(experiment_id):
experiment = updateCurrentExperiment(experiment_id)
clustering = ClusteringExp.from_json(experiment.output_dir())
num_clusters = clustering.num_clusters
num_instances_v = []
labels = []
for c in range(num_clusters):
instances_in_cluster = clustering.clusters[c].instances_ids
num_instances = len(instances_in_cluster)
# the empty clusters are not displayed
# if num_instances > 0:
num_instances_v.append(num_instances)
#labels.append('c_' + str(c))
labels.append(clustering.clusters[c].label)
barplot = BarPlot(labels)
dataset = PlotDataset(num_instances_v, 'Num. Instances')
barplot.add_dataset(dataset)
return jsonify(barplot.to_json())
def getTopWeightedFeatures(exp_id, inst_exp_id, instance_id, size, fold_id):
if fold_id == 'all':
return None
instance_id = int(instance_id)
exp = updateCurrentExperiment(exp_id)
inst_exp = updateCurrentExperiment(inst_exp_id)
# get the features
features_from_exp = FeaturesFromExp(inst_exp)
features_names, features_values = features_from_exp.get_instance(
instance_id)
features_values = [float(value) for value in features_values]
# get the pipeline with scaler and logistic model
experiment_dir = exp.output_dir()
if fold_id != 'None':
experiment_dir = path.join(experiment_dir, fold_id)
pipeline = joblib.load(path.join(experiment_dir, 'model', 'model.out'))
# scale the features
scaled_values = pipeline.named_steps['scaler'].transform(
np.reshape(features_values, (1, -1)))
weighted_values = np.multiply(scaled_values,
pipeline.named_steps['model'].coef_)
features = list(
map(lambda name, value, w_value: (name, value, w_value),
features_names, features_values, weighted_values[0]))
features.sort(key=lambda tup: abs(tup[2]))
features = features[:-int(size) - 1:-1]
features_names = [x[0] for x in features]
features_values = [x[1] for x in features]
features_weighted_values = [x[2] for x in features]
labels = [str(name) for name in features_names]
tooltips = [
'%s (%.2f)' % (name, features_values[i])
for i, name in enumerate(features_names)
]
barplot = BarPlot(labels)
dataset = PlotDataset(features_weighted_values, None)
dataset.set_color(colors_tools.red)
barplot.add_dataset(dataset)
return jsonify(barplot.to_json(tooltip_data=tooltips))
def display(self, directory):
labels = [
'0-10%', '10-20%', '20-30%', '30-40%', '40-50%', '50-60%',
'60-70%', '70-80%', '80-90%', '90-100%'
]
barplot = BarPlot(labels)
if not self.has_ground_truth:
dataset = PlotDataset(list(map(len, self.ranges)), 'numInstances')
dataset.set_color(colors_tools.get_label_color('all'))
barplot.add_dataset(dataset)
else:
self.displayLabel(barplot, labels_tools.MALICIOUS)
self.displayLabel(barplot, labels_tools.BENIGN)
filename = path.join(directory, 'predictions_barplot.json')
with open(filename, 'w') as f:
barplot.export_json(f)
class FeaturePlots(object):
def __init__(self, instances, feature_index):
self.feature_index = feature_index
features = instances.features
self.feature_type = features.types[self.feature_index]
self.feature_name = features.names[self.feature_index]
self.feature_id = features.ids[self.feature_index]
self.all_values = features.getValuesFromIndex(self.feature_index)
self._gen_plot_datasets(instances)
def compute(self):
if self.feature_type == FeatureType.binary:
self._gen_binary_histogram()
elif self.feature_type == FeatureType.numeric:
self._gen_bloxplot()
self._gen_histogram()
self._gen_density()
def export(self, output_dir):
output_dir = path.join(output_dir, str(self.feature_id))
dir_tools.createDirectory(output_dir)
if self.feature_type == FeatureType.binary:
with open(path.join(output_dir, 'binary_histogram.json'), 'w') as f:
self.barplot.export_json(f)
elif self.feature_type == FeatureType.numeric:
self.boxplot.display(path.join(output_dir, 'boxplot.png'))
with open(path.join(output_dir, 'histogram.json'), 'w') as f:
self.barplot.export_json(f)
self.density.display(path.join(output_dir, 'density.png'))
def _gen_plot_datasets(self, instances):
self.plot_datasets = {}
self._gen_label_plot_dataset(instances, labels_tools.MALICIOUS)
self._gen_label_plot_dataset(instances, labels_tools.BENIGN)
self._gen_label_plot_dataset(instances, 'unlabeled')
def _gen_label_plot_dataset(self, instances, label):
if label != 'unlabeled':
instances = instances.getAnnotatedInstances(label=label)
else:
instances = instances.getUnlabeledInstances()
values = instances.features.getValuesFromIndex(self.feature_index)
dataset = PlotDataset(values, label)
dataset.set_color(colors_tools.get_label_color(label))
self.plot_datasets[label] = dataset
def _gen_bloxplot(self):
self.boxplot = BoxPlot(title='Feature %s' % self.feature_name)
for label, dataset in self.plot_datasets.items():
if len(dataset.values) > 0:
self.boxplot.add_dataset(dataset)
def _gen_histogram(self):
# 10 equal-width bins computed on all the data
_, bin_edges = np.histogram(self.all_values, bins=10, density=False)
x_labels = ['%.2f - %.2f' % (bin_edges[e], bin_edges[e+1])
for e in range(len(bin_edges) - 1)]
self.barplot = BarPlot(x_labels)
for label, dataset in self.plot_datasets.items():
if len(dataset.values) > 0:
hist, _ = np.histogram(dataset.values, bins=bin_edges,
density=False)
hist_dataset = PlotDataset(hist, label)
hist_dataset.set_color(dataset.color)
self.barplot.add_dataset(hist_dataset)
def _gen_binary_histogram(self):
self.barplot = BarPlot(['0', '1'])
for label, dataset in self.plot_datasets.items():
if len(dataset.values) > 0:
num_0 = sum(dataset.values == 0)
num_1 = sum(dataset.values == 1)
hist_dataset = PlotDataset([num_0, num_1], label)
hist_dataset.set_color(dataset.color)
self.barplot.add_dataset(hist_dataset)
def _gen_density(self):
self.density = Density(title='Feature %s' % self.feature_name)
for _, dataset in self.plot_datasets.items():
if len(dataset.values) > 0:
self.density.add_dataset(dataset)