def process_csv_file(file):
df = pd.read_csv(file)
output_dictionary = {'option_name': [], 'input': [], 'grid_count': [], 'beta': [], 'date': [], 'minimizer': []}
output_dt = pd.DataFrame(output_dictionary)
day_count = len(df)
for i in range(2, day_count - 2):
today = df.iat[i + 2, 0]
data_block = df.iloc[i:i + 3]
# data_block = data_block.apply(pd.to_numeric)
option_name = 'unknown'
if not data_block.isnull().values.any():
# Then the 3-day data is good for us
option_ask = data_block.iloc[:, 1].values
option_bid = data_block.iloc[:, 2].values
volatility = data_block.iloc[:, 4].values
stock_ask = math.ceil(data_block.iat[2, 5] * 100) / 100
stock_bid = math.floor(data_block.iat[2, 6] * 100) / 100
if stock_ask == stock_bid:
stock_ask += 0.01
input_data = ns.DataBlock(today=today, option_ask=option_ask, option_bid=option_bid,
volatility=volatility, stock_ask=stock_ask, stock_bid=stock_bid)
input_data.create_system(grid, beta)
res = input_data.solve()
row = {'option_name': option_name, 'input': str(input_data), 'grid_count': grid, 'beta': beta, 'date': today,
'minimizer': np.array2string(res.x, max_line_width=np.inf, threshold=np.inf)}
output_dt = output_dt.append(row, ignore_index=True)
print(f'Solved {option_name} on {today}')
else:
print(f'{option_name} on {today} is skipped due to insufficient data.')
filename = Path(file).stem
output_file = f'{output_folder}/{filename}_out.csv'
output_dt.to_csv(output_file)
def predict(file, grid_count, beta):
df = pd.read_csv(file)
output_dictionary = {
'option_name': [],
'input': [],
'grid_count': [],
'beta': [],
'date': [],
'estimates': [],
'real': []
}
output_dt = pd.DataFrame(output_dictionary)
day_count = len(df)
day_count = 20
for i in range(2, day_count - 2):
today = df['DATE'][i]
data_block = df.iloc[i - 2:i + 1]
if len(set(data_block['OPTION_NAME'])) != 1:
continue
else:
option_name = data_block['OPTION_NAME'].iat[0]
value_block = data_block[[
'EOD_OPTION_PRICE_ASK', 'EOD_OPTION_PRICE_BID', 'IVOL_LAST',
'EOD_UNDERLYING_PRICE_ASK', 'EOD_UNDERLYING_PRICE_BID'
]]
if not value_block.isnull().values.any():
# Then the 3-day data is good for us
option_ask = value_block['EOD_OPTION_PRICE_ASK'].values
option_bid = value_block['EOD_OPTION_PRICE_BID'].values
volatility = value_block['IVOL_LAST'].values
stock_ask = float(value_block['EOD_UNDERLYING_PRICE_ASK'].iat[2])
stock_bid = float(value_block['EOD_UNDERLYING_PRICE_BID'].iat[2])
input_data = ns.DataBlock(today=today,
option_ask=option_ask,
option_bid=option_bid,
volatility=volatility,
stock_ask=stock_ask,
stock_bid=stock_bid)
input_data.create_system(grid_count, beta)
res = input_data.solve()
m = grid_count
solution = res.x.reshape(
(m - 1, m - 2))[[math.ceil(m / 2 - 1), m - 2],
math.ceil((m - 2) / 2)]
real_future = []
for j in range(2):
future_price = df['EOD_OPTION_PRICE_LAST'][i + j]
if np.isnan(future_price):
future_price = np.mean(
df[['EOD_OPTION_PRICE_ASK',
'EOD_OPTION_PRICE_BID']].iloc[i + j, :])
real_future.append(future_price)
row = {
'option_name': option_name,
'input': input_data.data(),
'grid_count': grid_count,
'beta': beta,
'date': today,
'estimates': solution,
'real': real_future
}
output_dt = output_dt.append(row, ignore_index=True)
print(
f'Solved {option_name} on {today}, finished {i-1}/{day_count-4}'
)
else:
print(
f'{option_name} on {today} is skipped due to insufficient data; finished {i-1}/{day_count-4}'
)
filename = Path(file).stem
output_file = f'output/prediction/{filename}_prediction.csv'
output_dt.to_csv(output_file)
from model import num_solver as ns
import math
block = ns.DataBlock(today='10/19/2016',\
option_ask = [0.86, 0.86, 0.86],\
option_bid = [0.84, 0.85, 0.85],\
volatility = [39.456, 38.061, 37.096],\
stock_ask = 4.66,\
stock_bid = 4.65)
m = 15
beta = 0.01
block.create_system(m, beta)
result = block.solve()
solution = result.x.reshape((m - 1, m - 2))
print('Minimizer is:')
print(solution)
print('Estimates for 1tau and 2tau: ', solution[[math.ceil(m / 2 - 1), m - 2],
math.ceil((m - 2) / 2)])
def solve(i, df, namespace, output_lock, day_count):
if 'DATE' in df:
today = df['DATE'][i]
else:
today = None
data_block = df.iloc[i - 2:i + 1]
if 'OPTION_NAME' in df:
if len(set(data_block['OPTION_NAME'])) > 1:
return
else:
option_name = data_block['OPTION_NAME'].iat[0]
else:
option_name = None
value_block = data_block[value_list]
if not value_block.isnull().values.any():
# Then the 3-day data is good for us
option_ask = value_block['EOD_OPTION_PRICE_ASK'].values
option_bid = value_block['EOD_OPTION_PRICE_BID'].values
volatility = value_block['IVOL_LAST'].values
stock_ask = float(value_block['EOD_UNDERLYING_PRICE_ASK'].iat[2])
stock_bid = float(value_block['EOD_UNDERLYING_PRICE_BID'].iat[2])
input_data = ns.DataBlock(today=today,
option_ask=option_ask,
option_bid=option_bid,
volatility=volatility,
stock_ask=stock_ask,
stock_bid=stock_bid)
input_data.create_system(grid_count, beta)
res = input_data.solve()
m = grid_count
solution = res.x.reshape((m - 1, m - 2))[[math.ceil(m / 2 - 1), m - 2],
math.ceil((m - 2) / 2)]
real_future = []
for j in range(2):
if 'EOD_OPTION_PRICE_LAST' in df:
future_price = df['EOD_OPTION_PRICE_LAST'][i + j]
else:
future_price = np.mean(
df[['EOD_OPTION_PRICE_ASK',
'EOD_OPTION_PRICE_BID']].iloc[i + j, :])
real_future.append(future_price)
row = {
'option_name': option_name,
'grid_count': grid_count,
'beta': beta,
'date': today,
'option_ask-2': input_data.u_a_list[0],
'option_ask-1': input_data.u_a_list[1],
'option_ask0': input_data.u_a_list[2],
'option_bid-2': input_data.u_b_list[0],
'option_bid-1': input_data.u_b_list[1],
'option_bid0': input_data.u_b_list[2],
'stock_ask0': input_data.s_a,
'stock_bid0': input_data.s_b,
'ivol-2': input_data.ivol_list[0],
'ivol-1': input_data.ivol_list[1],
'ivol-0': input_data.ivol_list[2],
'est+1': float(solution[0]),
'est+2': float(solution[1]),
'real+1': float(real_future[0]),
'real+2': float(real_future[1])
}
output_lock.acquire()
try:
namespace.df = namespace.df.append(row, ignore_index=True)
print(
f'Solved {option_name} on {today}, finished {i - 1}/{day_count - 4}'
)
finally:
output_lock.release()
else:
print(
f'{option_name} on {today} is skipped due to insufficient data; finished {i - 1}/{day_count - 4}'
)