def find_principal_securities(marketcode, event_filename, date_range, T,
filename):
event_data = np.loadtxt('csv/' + event_filename + '.csv',
dtype=np.unicode,
delimiter=',')[1:, :]
date_from = datetime.strptime(date_range[0], base.DATE_FORMAT)
date_to = datetime.strptime(date_range[1], base.DATE_FORMAT)
event_date_list = []
for event in event_data:
date = datetime.strptime(event[1], base.DATE_FORMAT)
if date_from <= date <= date_to:
event_date_list.append(date.date())
offset = 0
num_clusters = 4
event_security_cluster = []
change_generator, security_list = base.observed_changes(marketcode,
date_range,
always=True,
T=T)
event_idx = 0
event_date = event_date_list[event_idx]
last_data_change = np.array([])
for data_change, date in change_generator:
if date > event_date:
traded_securities_ids = ~np.all(last_data_change == 0, axis=0)
effective_change = last_data_change[:, traded_securities_ids]
traded_securities = security_list[traded_securities_ids]
corr = pearson_correlation(effective_change, offset)
cluster_idx = cluster_data(corr, num_clusters=num_clusters)
security_clusters = np.empty(len(security_list)) * np.nan
for i in range(num_clusters):
security_clusters[np.in1d(
security_list[:, 0], traded_securities[cluster_idx == i,
0])] = i
event_security_cluster.append(security_clusters)
event_idx = event_idx + 1
if event_idx >= len(event_date_list):
break
event_date = event_date_list[event_idx]
last_data_change = data_change
event_security_cluster = np.array(event_security_cluster)
header = ['ISIN', 'issuer']
data_csv = np.array([header])
for row in security_list[np.all(event_security_cluster == 0, axis=0)]:
data_csv = np.append(data_csv, [row], axis=0)
np.savetxt('csv/' + filename + '.csv', data_csv, fmt='%s', delimiter=',')
def market_stress_matrix(marketcode, date, T, filename):
num_clusters = 4
change_generator, security_list = base.observed_changes(marketcode, (date,date), T=T)
change_data, date = next(change_generator)
corr = pearson_correlation(change_data)
order_metric = np.average(corr, axis=0)
cluster_idx = cluster_data(corr, num_clusters=num_clusters)
data_sorted_idx = np.lexsort([np.negative(order_metric), cluster_idx])
corr = corr[data_sorted_idx,:][:,data_sorted_idx]
security_list = security_list[data_sorted_idx]
header = ['isin'] + [security[0] for security in security_list]
data_csv = np.array([header])
for security_idx, security in enumerate(security_list):
data_csv = np.append(data_csv, np.concatenate(([[security[0]]], [corr[security_idx]]), axis=1), axis=0)
np.savetxt('csv/'+filename+'.csv', data_csv, fmt='%s', delimiter=',')
param_fit_normal, param_error_normal = base.plot_baseline_stress(
baseline_stress_file)
######
market_code = 'SWX'
series_length = 20
event_filename = 'market_stress_events'
date_range = ('01.01.2000', '31.12.2018')
stress_series_file = 'stress_series_{}_{}_{}-{}'.format(
market_code, series_length, base.datestamp(date_range[0]),
base.datestamp(date_range[1]))
if recalculate:
change_generator, security_list = base.observed_changes(market_code,
date_range,
always=False,
T=series_length)
code.stress_series(change_generator, offset=0, filename=stress_series_file)
else:
security_list = [None] * 314
base.plot_time_series({'T=20': stress_series_file},
event_filename=event_filename)
data_range_zoom = ('01.01.2016', '31.12.2016')
stress_series_zoom_file = 'stress_series_{}_{}_{}-{}'.format(
market_code, series_length, base.datestamp(data_range_zoom[0]),
base.datestamp(data_range_zoom[1]))
if recalculate:
change_generator, _ = base.observed_changes(market_code,
data_range_zoom,
always=False,