def generate_graph(num) -> Graph:
graph = Graph(f'Graph_{num}')
# product_count = random.randint(2, 10)
product_count = 1000
for n in range(product_count):
price = random.randint(5, 500)
product_node = graph.add_node({
'type': 'PRODUCT',
'price': price
},
f'Product_{n}',
node_type='PRODUCT')
if with_probability(0.04):
new_price_rate = rand_float(0.65, 0.8)
change_node = graph.add_node(
{
'type': 'CHANGE',
'old_price': price,
'new_price': price * new_price_rate,
'is_fraud': True
},
f'Change_of_Prod_{n}',
node_type='CHANGE',
node_color='red')
change_node.add_neighbor(product_node)
elif with_probability(0.45):
new_price_rate = rand_float(0.8, 1)
change_node = graph.add_node(
{
'type': 'CHANGE',
'old_price': price,
'new_price': price * new_price_rate,
'is_fraud': False
},
f'Change_of_Prod_{n}',
node_type='CHANGE',
node_color='green')
change_node.add_neighbor(product_node)
# for node in graph._nodes:
# link_count = random.randint(0, min(8, int(node_count / 4)))
#
# for _ in range(link_count):
# node.add_neighbor(graph.get_random_node())
return graph
def generate_graph() -> Tuple[Graph, pd.Series]:
graph = Graph(f'Graph Node Level')
node_count = 25_000
nodes_gt = pd.Series()
for n in range(node_count):
type = random.choice(['type-1', 'type-2', 'type-3'])
node = graph.add_node({}, f'Node_{n}', node_type=type)
nodes_gt._set_value(node.get_id(), type)
for node in graph._nodes:
link_count = random.randint(0, 2) #random.randint(0, min(7, int(node_count / 4)))
for _ in range(link_count):
node.add_neighbor(graph.get_random_node())
return graph, nodes_gt
def generate_graph(self, db: Database) -> Graph:
print(f'generating graph for {db.get_name()}')
graph = Graph(name=db.get_name())
# generate nodes
# time_start = time.perf_counter()
for table in db.get_tables():
# print(f'adding nodes for table {table.get_name()}')
table_data = table.get_data()
for index, row in table_data.iterrows():
properties = {
column.get_name(): row[column.get_name()]
for column in table.get_columns().values() if
not column.get_is_primary() and not column.get_is_hidden()
}
graph.add_node(
properties,
f'{table.get_name()}[{index}]',
node_color=self._fraud_node_color if 'is_fraud' in row
and row['is_fraud'] else self._default_node_color,
node_type=table.get_name())
# time_end = time.perf_counter()
# print(f"generate nodes took {time_end - time_start:0.4f} seconds")
# generate DELETE nodes from MTA_CHANGES
# time_start = time.perf_counter()
for _, src_record in db.get_table('MTA_CHANGES').get_data().iterrows():
if (src_record['change_type']) == 'delete':
dst_table_name, _ = src_record['table_column_ref'].split(
'.', 2)
graph.add_node(src_record['old_value'],
f'{dst_table_name}[{src_record["record_id"]}]',
self._fraud_node_color if src_record['is_fraud']
else self._default_node_color,
node_type='MTA_CHANGES')
# time_end = time.perf_counter()
# print(f"delete nodes took {time_end - time_start:0.4f} seconds")
# generate links
# time_start = time.perf_counter()
for table in db.get_tables():
# print(f'TABLE KEYS: {table.get_name()}')
foreign_keys = table.get_all_foreign_keys()
table_data = table.get_data()
for index, src_record in table_data.iterrows():
src_node = graph.get_node_by_key(
f'{table.get_name()}[{index}]')
for foreign_key in foreign_keys:
if foreign_key.get_src_table() != table.get_name():
continue
dst_table = db.get_table(foreign_key.get_dst_table())
dst_data_entry_index = dst_table.find_record_index(
foreign_key.get_dst_column(),
src_record[foreign_key.get_src_column()])
dst_node = graph.get_node_by_key(
f'{foreign_key.get_dst_table()}[{dst_data_entry_index}]'
)
if dst_node is not None:
if foreign_key.get_reverse_relation():
dst_node.add_neighbor(src_node,
foreign_key.get_color())
else:
src_node.add_neighbor(dst_node,
foreign_key.get_color())
# time_end = time.perf_counter()
# print(f"generate links took {time_end - time_start:0.4f} seconds")
# generate meta links
# time_start = time.perf_counter()
for index, src_record in db.get_table(
'MTA_CHANGES').get_data().iterrows():
src_node = graph.get_node_by_key(f'MTA_CHANGES[{index}]')
dst_table_name, dst_column_name = src_record[
'table_column_ref'].split('.', 2)
dst_node = graph.get_node_by_key(
f'{dst_table_name}[{src_record["record_id"]}]')
src_node.add_neighbor(dst_node,
self._edge_colors[src_record['change_type']])
# time_end = time.perf_counter()
# print(f"generate meta links took {time_end - time_start:0.4f} seconds")
return graph
