def _all_candlestick_patterns(self, df):
h = df.High.values
o = df.Open.values
l = df.Low.values
c = df.Close.values
v = df.Volume.values
name = 'CDL2CROWS'
df[name] = talib.CDL2CROWS(o,h,l,c)
talib.get_function_groups()
return df
def candle_analyse(df):
'''
input:OHLC dataframe
'''
cn_names = []
## calc all candle score
open, high, low, close = df['open'], df['high'], df['low'], df['close']
df = df[['date']].copy()
func_names = talib.get_function_groups()['Pattern Recognition']
for func_name in func_names:
func = getattr(talib, func_name)
score_data = func(open, high, low, close)
patt_name = func_name[3:]
# if score_data!=0:
df[patt_name] = score_data
cn_names.append(patt_name)
def make_dict(row):
res_d = OrderedDict()
for func_name in func_names:
patt_name = func_name[3:]
score = row[patt_name]
if score != 0:
res_d[patt_name] = score
st = ','.join(['%s:%s' % (k, v) for k, v in res_d.items()])
return st
df['CDLList'] = df.apply(make_dict, axis=1)
###
total_cdl_score = df[cn_names].sum(axis=1)
df['CDLScore'] = total_cdl_score
# print jsdump(cdl_info)
return df
def getTriggeredPatternsPrediction(stockData, date, timeGap=1):
patternFuncs = talib.get_function_groups()['Pattern Recognition']
patternAnalysis = {}
patternAccuracy = {}
for patternFunc in patternFuncs:
patternAnalyzer = getattr(talib, patternFunc)
predictionInfo = patternAnalyzer(stockData['Open'], stockData['High'],
stockData['Low'], stockData['Close'])
try:
predictionInfo[stockData[stockData['Date'] == date].index[0]]
except:
print('The date given is a non-trading day (i.e. Weekend)')
return pd.DataFrame({})
if (predictionInfo[stockData[stockData['Date'] == date].index[0]] !=
0):
print(patternFunc)
print('Accuracy:',
patternPrediction(stockData, predictionInfo, timeGap))
patternAccuracy[patternFunc] = patternPrediction(
stockData, predictionInfo, timeGap)
# newDF = stockData
# newDF['Action'] = predictionInfo
# newDF.to_csv(patternFunc+'.csv', index=False)
patternAnalysis[patternFunc] = predictionInfo.tail(1)
patternAnalysis = pd.DataFrame(patternAnalysis)
return pd.DataFrame(patternAccuracy)
def patternDetection(historical_klines):
op = []
hi = []
lo = []
cl = []
# collecting all open, high, low and close values
for kline in historical_klines:
op.append(float(kline[1]))
hi.append(float(kline[2]))
lo.append(float(kline[3]))
cl.append(float(kline[4]))
# converting lists into (talib-compatible) numpy arrays
op = numpy.array(op, dtype=float)
hi = numpy.array(hi, dtype=float)
lo = numpy.array(lo, dtype=float)
cl = numpy.array(cl, dtype=float)
# applying the talib candle functions for pattern recognition on course data
candle_names = talib.get_function_groups()['Pattern Recognition']
for candle in candle_names:
results = getattr(talib, candle)(op, hi, lo, cl)
for i in range(len(results)):
# '0' means that no pattern was found
if(results[i] != 0):
dateOfMatch = datetime.today() - timedelta(days=i+1)
# collect data of the match and print it
print("Pattern: " + candle + " Result: " + str(results[i]) + " - " + str(dateOfMatch))
def onClickedFunctionButton(self):
loader = QUiLoader()
function_ui = loader.load('function_description.ui',
parentWidget=self.python_editor)
function_ui.setWindowTitle("函数说明")
function_ui.show()
self.tree = function_ui.findChild(QTreeWidget)
self.demo = function_ui.findChild(QTextEdit)
talib_groups = talib.get_function_groups()
for key in talib_groups.keys():
node = QTreeWidgetItem(self.tree)
node.setText(0, key)
for value in talib_groups[key]:
sub_node = QTreeWidgetItem(node)
sub_node.setText(0, value)
names = functions.functions_name
for key in names.keys():
node = QTreeWidgetItem(self.tree)
node.setText(0, key)
for name in names[key].keys():
sub_node = QTreeWidgetItem(node)
sub_node.setText(0, name)
self.tree.itemClicked.connect(
lambda event1, event2: self.onFunctionTreeItemClicked(
event1, event2))
self.tree.itemDoubleClicked.connect(
lambda event1, event2: self.onFunctionTreeItemDoubleClicked(
event1, event2))
return
def get_ta_functions():
# func_names = []
func_groups = ['Volume Indicators', 'Volatility Indicators',
'Overlap Studies', 'Momentum Indicators']
func_names = [func_name for g in func_groups for func_name in ta.get_function_groups()[g]]
funcs = {func_name: Function(func_name) for func_name in func_names}
return funcs
def featurize_data(data):
'''
Takes in a CSV with the following columns:
=======================================
Date,Open,High,Low,Close,Volume,OpenInt
=======================================
and assigns a label (-1 or 1) for each price point given whether the price
point is higher or lower than the previous one.
'''
# Initialize output array
output = np.ones((len(data), 1))
# Iterate over function to match candlestick patterns
for func in talib.get_function_groups()['Pattern Recognition']:
# Interpret strings as functions and call them on numeric data
func_to_call = getattr(talib, func)
feature_column = func_to_call(data['Open'], data['High'], data['Low'],
data['Close']).reshape(len(data), 1)
# Column join indicator output for candlestick pattern features
output = np.concatenate((output, feature_column), axis=1)
# Concatenate labels to final dataframe
output = np.concatenate((output, data['Label'].reshape(len(data), 1)),
axis=1)
output = output[:, 1:]
return output
def get_cs_column_names():
"""
gets candlestick pattern column names for df
"""
pattern_functions = talib.get_function_groups()['Pattern Recognition']
column_names = [p[3:]
for p in pattern_functions] # for dataframe; cuts out CDL
return column_names
def get_function_groups():
'''
Returns a dict with keys of function-group names and values of lists
of function names ie {'group_names': ['function_names']}
返回带分组的函数列表
'''
return talib.get_function_groups()
def ohlc_to_ta_lib(data):
ta_lib_dict = {'bull': [], 'bear': []}
#다른 형태의 api제공을 위해 가공
ta_lib_dict_type2 = {}
csv_string = data
csv_data = pd.read_csv(io.StringIO(csv_string),
index_col=0,
parse_dates=True)
patterns_names = talib.get_function_groups()['Pattern Recognition']
#ta-lib pattern recognition list print(patterns_names)
for pattern in patterns_names:
csv_data[pattern] = getattr(talib,
pattern)(csv_data['Open'],
csv_data['High'], csv_data['Low'],
csv_data['Close'])
tmp_data = csv_data
#pattern이 없는 0칼럼들 다 제외
zero_column = list(csv_data.columns[(csv_data == 0).all()])
not_zero_column = list(set(patterns_names) - set(zero_column))
#print(not_zero_column)
#csv_data.to_csv('./sample.csv', sep=',', na_rep='NaN')
for pattern in not_zero_column:
csv_data_bull = tmp_data[tmp_data[pattern] > 0]
csv_data_bear = tmp_data[tmp_data[pattern] < 0]
#각 csv_data에서 양수 인거(보통 100)은 bullish를 나타내고 음수 인거(보통 -100)은 bearish를 나타낸다.
# 각 패턴이 bull version도 있고 bear 버전도 있다. 따라서 나눠야 한다.
date_bull = list(
map(lambda x: x.strftime("%Y-%m-%d"), csv_data_bull.index))
date_bear = list(
map(lambda x: x.strftime("%Y-%m-%d"), csv_data_bear.index))
#print(date_bull)
for date in date_bull:
pattern_name = f'{pattern[3:]}_BULL'
ta_lib_dict['bull'].append((pattern_name, date))
#다른 형태의 api제공을 위해 가공
if pattern_name not in ta_lib_dict_type2.keys():
ta_lib_dict_type2[pattern_name] = [(date, "bull")]
else:
ta_lib_dict_type2[pattern_name].append((date, "bull"))
for date in date_bear:
pattern_name = f'{pattern[3:]}_BEAR'
ta_lib_dict['bear'].append((pattern_name, date))
#다른 형태의 api제공을 위해 가공
if pattern_name not in ta_lib_dict_type2.keys():
ta_lib_dict_type2[pattern_name] = [(date, "bear")]
else:
ta_lib_dict_type2[pattern_name].append((date, "bear"))
return (ta_lib_dict, ta_lib_dict_type2, tmp_data)
def list_indicator(group=None):
""" list all available indicators
Parameters:
-------
group: string (optional)
group of indicator; if none, function will list all indicators
"""
if group is None:
return talib.get_functions()
else:
if group in ["Overlap", "overlap", "over"]:
return talib.get_function_groups()["Overlap Studies"]
elif group is ["Momentum", "momentum", "momen", "mom"]:
return talib.get_function_groups()["Momentum Indicators"]
elif group is ["Volume", "volume", "volu", "vol"]:
return talib.get_function_groups()["Volume Indicators"]
elif group is ["Volatility", "volatility", "vola"]:
return talib.get_function_groups()["Volatility Indicators"]
elif group is ["Price", "price", "pri"]:
return talib.get_function_groups()["Price Transform"]
elif group is ["Cycle", "cycle", "cyc"]:
return talib.get_function_groups()["Cycle Indicators"]
elif group is ["Pattern", "pattern", "patt", "pat"]:
return talib.get_function_groups()["Pattern Recognition"]
else:
print "ERROR: no matching group " + group + ", must be in: "
print list_indicator_groups()
print "Or enter no group to get full list of indicators:"
print " list_indicator()"
return
def list_indicator(group=None):
""" list all available indicators
Parameters:
-------
group: string (optional)
group of indicator; if none, function will list all indicators
"""
if group is None:
return talib.get_functions()
else:
if group in ["Overlap", "overlap", "over"]:
return talib.get_function_groups()["Overlap Studies"]
elif group is ["Momentum", "momentum", "momen", "mom"]:
return talib.get_function_groups()["Momentum Indicators"]
elif group is ["Volume", "volume", "volu", "vol"]:
return talib.get_function_groups()["Volume Indicators"]
elif group is ["Volatility", "volatility", "vola"]:
return talib.get_function_groups()["Volatility Indicators"]
elif group is ["Price", "price", "pri"]:
return talib.get_function_groups()["Price Transform"]
elif group is ["Cycle", "cycle", "cyc"]:
return talib.get_function_groups()["Cycle Indicators"]
elif group is ["Pattern", "pattern", "patt", "pat"]:
return talib.get_function_groups()["Pattern Recognition"]
else:
print "ERROR: no matching group "+group+", must be in: "
print list_indicator_groups()
print "Or enter no group to get full list of indicators:"
print " list_indicator()"
return
def get_indicators_by_group(group: str) -> [(str, str)]:
"""Returns list of select options containing abbreviations of the groups indicators"""
indicator_selects = []
group_indicators = ta.get_function_groups()[group]
for i in range(len(group_indicators)):
abbrev = group_indicators[i]
func = getattr(ab, abbrev)
name = func.info["display_name"]
indicator_selects.append((abbrev, abbrev))
return indicator_selects
def pattern_chandeliers(iDf, iPatternList=talib.get_function_groups()['Pattern Recognition']):
"""
:param iDf:
:param iPatternList:
:return:
"""
for pattern in iPatternList:
func = getattr(talib.abstract, pattern)
iDf.loc[:, pattern] = func(iDf)
return iDf
def onFunctionTecTree(self, item, column, function_ui, data,
hidden_columns):
group_box = function_ui.findChild(QGroupBox, "parameter_box")
display_name_box = function_ui.findChild(QLineEdit, "display_name")
labels = group_box.findChildren(QLabel)
spin_boxs = group_box.findChildren(QSpinBox)
for label in labels:
label.close()
for spin_box in spin_boxs:
spin_box.close()
text = item.text(column)
if text in talib.get_function_groups().keys():
return
indicator = abstract.Function(text.lower())
info = indicator.info
display_name = info.get("display_name", "")
display_name_box.setText(display_name)
params_dict = indicator.get_parameters()
keys = list(params_dict.keys())
keys.insert(0, u"Column")
locator = 20
num = 0
paras_locators = [locator + i * 30 for i in range(len(keys))]
for key in keys:
p = paras_locators[num]
label = QLabel(group_box)
label.setText(key)
label.setGeometry(10, p, 100, 20)
if key == "Column":
combo_box = QComboBox(group_box)
combo_box.setGeometry(100, p, 100, 20)
combo_box.setObjectName(key)
combo_box.setWhatsThis(key)
columns = [
i for i in list(data.columns) if i not in hidden_columns
]
combo_box.addItems(columns)
if "close" in columns:
combo_box.setCurrentText("close")
combo_box.show()
else:
spin_box = QSpinBox(group_box)
spin_box.setGeometry(100, p, 50, 20)
spin_box.setObjectName(text)
spin_box.setWhatsThis(key)
value = params_dict[key]
spin_box.setValue(value)
spin_box.show()
label.show()
num += 1
def feature_dfs():
mom_ind = talib.get_function_groups()['Momentum Indicators']
over_stud = talib.get_function_groups()['Overlap Studies']
volu_ind = talib.get_function_groups()['Volume Indicators']
cyc_ind = talib.get_function_groups()['Cycle Indicators']
vola_ind = talib.get_function_groups()['Volatility Indicators']
stats_ind = talib.get_function_groups()['Statistic Functions']
talib_abstract_fun_list = mom_ind + over_stud + volu_ind + cyc_ind + vola_ind + stats_ind
talib_abstract_fun_list.remove('MAVP')
no_params_df = pd.DataFrame([])
only_time_period_df = pd.DataFrame([])
other_param_df = pd.DataFrame([])
for fun in talib_abstract_fun_list:
info = getattr(talib.abstract, fun).info
data = pd.Series([
info['group'], info['name'], info['display_name'],
[
'{}: {}'.format(key, value)
for key, value in info['parameters'].items()
], info['output_names']
])
if len(info['parameters']) == 0:
no_params_df = no_params_df.append(data, ignore_index=True)
elif 'timeperiod' in info['parameters'] and len(
info['parameters']) == 1:
only_time_period_df = only_time_period_df.append(data,
ignore_index=True)
else:
other_param_df = other_param_df.append(data, ignore_index=True)
ind_dfs = [no_params_df, only_time_period_df, other_param_df]
for ind_df in ind_dfs:
ind_df.columns = [
'Group', 'Name', 'Short Description', 'Parameters', 'Output Names'
]
return no_params_df, only_time_period_df, other_param_df
def add_TA(df: pd.DataFrame):
inputs = {
'open': df["Open"],
'high': df["High"],
'low': df["Low"],
'close': df["Close"],
'volume': df["Volume"]
}
overlap_studies = talib.get_function_groups()['Overlap Studies']
if "MAVP" in overlap_studies:
overlap_studies.remove("MAVP")
df = process_ta_functions_group(df, inputs, overlap_studies)
df = process_ta_functions_group(
df, inputs,
talib.get_function_groups()['Momentum Indicators'])
df = process_ta_functions_group(
df, inputs,
talib.get_function_groups()['Cycle Indicators'])
for func in talib.get_function_groups()['Volume Indicators']:
df[func] = globals()[func](inputs)
for func in talib.get_function_groups()['Volatility Indicators']:
df[func] = globals()[func](inputs)
for func in talib.get_function_groups()['Pattern Recognition']:
df[func] = globals()[func](inputs)
def get_price_patterns(data):
"""
Detects common price patterns using TA-lib, e.g. Two Crows, Belt-hold, Hanging Man etc.
:param data: DataFrame with data.
:return: DataFrame with pattern "likelihoods" on -200 - 200 scale.
"""
patterns = {
name: getattr(talib, name)(data['Open Price'], data['High Price'],
data['Low Price'], data['Close Price'])
for name in talib.get_function_groups()['Pattern Recognition']
}
return pd.DataFrame(patterns)
def onLoadFunctionDialogTab(self, index, function_ui, data,
hidden_columns):
if index == 0:
self.cal_list1 = function_ui.findChild(QListWidget, "cal_list1")
self.cal_list2 = function_ui.findChild(QListWidget, "cal_list2")
self.function_box = function_ui.findChild(QComboBox,
"function_box")
self.list_items = [
i for i in list(data.columns) if i not in hidden_columns
]
self.cal_list1.clear()
self.cal_list2.clear()
for item in self.list_items:
self.cal_list1.addItem(item)
for item in self.list_items:
self.cal_list2.addItem(item)
self.function_box.currentIndexChanged.connect(
self.onFunctionCalculateBox)
elif index == 1:
self.rel_list1 = function_ui.findChild(QListWidget, "rel_list1")
self.rel_list2 = function_ui.findChild(QListWidget, "rel_list2")
self.function_box = function_ui.findChild(QComboBox, "rel_box")
self.list_items = [
i for i in list(data.columns) if i not in hidden_columns
]
self.rel_list1.clear()
self.rel_list2.clear()
for item in self.list_items:
self.rel_list1.addItem(item)
for item in self.list_items:
self.rel_list2.addItem(item)
self.function_box.currentIndexChanged.connect(
self.onFunctionCalculateBox)
elif index == 2:
self.search_tec_function = function_ui.findChild(
QLineEdit, "search_input")
self.tec_tree = function_ui.findChild(QTreeWidget, "tec_tree")
talib_groups = talib.get_function_groups()
for key in talib_groups.keys():
node = QTreeWidgetItem(self.tec_tree)
node.setText(0, key)
for value in talib_groups[key]:
sub_node = QTreeWidgetItem(node)
sub_node.setText(0, value)
self.tec_tree.itemClicked.connect(
lambda event1, event2: self.onFunctionTecTree(
event1, event2, function_ui, data, hidden_columns))
self.search_tec_function.textEdited.connect(
lambda event: self.onTecFunctionSearched(function_ui))
class CDL_Patterns(CustomFactor):
"""
模式识别（所有模式，多输出）
.. code-block:: python
cdls = CDL_Patterns()
a = cdls.CDL2CROWS
b = cdls.CDL3BLACKCROWS
returns
-------
talib所有模式值Integer
"""
window_length = 5 # 模式识别窗口最少为1，最多5。取5?
inputs = [
USEquityPricing.open, USEquityPricing.high, USEquityPricing.low,
USEquityPricing.close
]
mask = IsStock()
outputs = ta.get_function_groups()['Pattern Recognition']
window_safe = True
def _validate(self):
super(CDL_Patterns, self)._validate()
if self.window_length < 5:
raise ValueError(
"For pattern recognition,'CDL_Patterns' expected a window length of at least 5, but was "
"given {window_length}. ".format(
window_length=self.window_length))
def _one_pattern(self, out, func_name, opens, highs, lows, closes, N):
"""计算指定模式名称因子"""
p_func = ta.abstract.Function(func_name)
res = np.zeros(N)
for i in range(N):
inputs = {
'open': opens[:, i],
'high': highs[:, i],
'low': lows[:, i],
'close': closes[:, i]
}
p_func.input_arrays = inputs
res[i] = p_func.outputs[-1]
setattr(out, func_name, res)
def compute(self, today, assets, out, opens, highs, lows, closes):
N = len(assets)
for func_name in self.outputs:
self._one_pattern(out, func_name, opens, highs, lows, closes, N)
def get_groups_markdown(update=False):
def unpluralize(noun):
if noun.endswith('s'):
if len(noun) > 2 and noun[-2] not in ["'", 'e']:
return noun[:-1]
return noun
doc_links = get_doc_links(update)
ret = {}
for group, funcs in talib.get_function_groups().items():
h1 = '# %s' % unpluralize(group)
h1 = h1 + ' Functions' if 'Function' not in h1 else h1 + 's'
ret[group] = [h1]
for func in funcs:
# figure out this function's options
f = Function(func)
inputs = f.info['input_names']
if 'prices' in inputs:
input_names = ', '.join(inputs['prices'])
else:
input_names = ', '.join([x for x in inputs.values() if x])
params = ', '.join(
['%s=%i' % (param, default)
for param, default in f.info['parameters'].items()])
outputs = ', '.join(f.info['output_names'])
# print the header
ret[group].append('### %s - %s' % (func, f.info['display_name']))
# print the code definition block
ret[group].append("```")
if params:
ret[group].append('%s = %s(%s, %s)' % (
outputs, func.upper(), input_names, params))
else:
ret[group].append('%s = %s(%s)' % (
outputs, func.upper(), input_names))
ret[group].append("```\n")
# print extra info if we can
if func in doc_links:
ret[group].append(
'Learn more about the %s at [tadoc.org](%s). ' % (
f.info['display_name'], doc_links[func]))
ret[group].append('\n[Documentation Index](../doc_index.html)')
ret[group].append('[FLOAT_RIGHTAll Function Groups](../funcs.html)')
ret[group] = '\n'.join(ret[group]) + '\n'
return ret
def get_groups_markdown(update=False):
def unpluralize(noun):
if noun.endswith('s'):
if len(noun) > 2 and noun[-2] not in ["'", 'e']:
return noun[:-1]
return noun
doc_links = get_doc_links(update)
ret = {}
for group, funcs in talib.get_function_groups().items():
h1 = '# %s' % unpluralize(group)
h1 = h1 + ' Functions' if 'Function' not in h1 else h1 + 's'
ret[group] = [h1]
for func in funcs:
# figure out this function's options
f = Function(func)
inputs = f.info['input_names']
if 'prices' in inputs:
input_names = ', '.join(inputs['prices'])
else:
input_names = ', '.join([x for x in inputs.values() if x])
params = ', '.join([
'%s=%i' % (param, default)
for param, default in f.info['parameters'].items()
])
outputs = ', '.join(f.info['output_names'])
# print the header
ret[group].append('### %s - %s' % (func, f.info['display_name']))
# print the code definition block
ret[group].append("```")
if params:
ret[group].append('%s = %s(%s, %s)' %
(outputs, func.upper(), input_names, params))
else:
ret[group].append('%s = %s(%s)' %
(outputs, func.upper(), input_names))
ret[group].append("```\n")
# print extra info if we can
if func in doc_links:
ret[group].append(
'Learn more about the %s at [tadoc.org](%s). ' %
(f.info['display_name'], doc_links[func]))
ret[group].append('\n[Documentation Index](../doc_index.html)')
ret[group].append('[FLOAT_RIGHTAll Function Groups](../funcs.html)')
ret[group] = '\n'.join(ret[group]) + '\n'
return ret
def csv_to_func(ticker, dataset_name):
try:
#base_path = f'./pattern_datasets/{dataset_name}'
base_path = f'./pattern_datasets/{dataset_name}/Industrials'
if not os.path.isdir(base_path):
os.makedirs(base_path, exist_ok=True)
patterns_names = talib.get_function_groups()['Pattern Recognition']
#print(patterns_names)
#csv_data = pd.read_csv(f"./datasets/daily/{ticker}",index_col=0,parse_dates=True)
csv_data = pd.read_csv(f"./datasets/Industrials/{ticker}",
index_col=0,
parse_dates=True)
cnt = 1
for pattern in patterns_names:
csv_data[pattern] = getattr(talib, pattern)(csv_data['Open'],
csv_data['High'],
csv_data['Low'],
csv_data['Close'])
tmp_data = csv_data
#해당 구간에서 한번도 패턴이 안나온 항목들을 제외한다.
zero_column = list(csv_data.columns[(csv_data == 0).all()])
not_zero_colum = list(set(patterns_names) - set(zero_column))
for pattern in not_zero_colum:
#print(pattern)
csv_data_bull = tmp_data[tmp_data[pattern] > 0]
csv_data_bear = tmp_data[tmp_data[pattern] < 0]
#각 csv_data에서 양수 인거(보통 100)은 bullish를 나타내고 음수 인거(보통 -100)은 bearish를 나타낸다.
# 각 패턴이 bull version도 있고 bear 버전도 있다. 따라서 나눠야 한다.
date_bull = list(
map(lambda x: x.strftime("%Y-%m-%d"), csv_data_bull.index))
date_bear = list(
map(lambda x: x.strftime("%Y-%m-%d"), csv_data_bear.index))
for idx, item in enumerate(date_bull):
make_chart(ticker, tmp_data, item, cnt, pattern, "bull",
base_path)
cnt += 1
for idx, item in enumerate(date_bear):
make_chart(ticker, tmp_data, item, cnt, pattern, "bear",
base_path)
cnt += 1
except Exception as ex:
print("error occured", ex)
pass
def pattern(ohlc):
inputs = {"open":np.array(ohlc.open.values),
"high": np.array(ohlc.high.values),
"low": np.array(ohlc.low.values),
"close": np.array(ohlc.close.values),
"volume": np.array(ohlc.volume.values).astype(float)}
dict1 = talib.get_function_groups()
pattern = dict1['Pattern Recognition']
dict2 = {}
dict1 ={}
for item in pattern:
value = eval("talib.{}(inputs['open'],inputs['high'], inputs['low'], inputs['close'])".format(item))
dict1["{}".format(item)] = value
return dict1
def pattern_pref(df, khh):
df = df[df["custid"] == khh]
stocks = np.unique(np.asarray(df["stkcode"]))
patterns = talib.get_function_groups()['Pattern Recognition']
pref_result = []
for pattern_index, pattern in enumerate(patterns):
op = [0, 0.0001]
pattern_type = pattern_dict[pattern_index]
if not pattern_type == 'discard':
for stock in stocks:
temp = df[df["stkcode"] == stock]
pat_op = pattern_op(temp, stock, pattern, pattern_type)
op[0] += pat_op[0]
op[1] += pat_op[1]
pref_result.append(op[0] / op[1])
return pref_result
def get_last_day_patterns(df):
candle_names = talib.get_function_groups()['Pattern Recognition']
patterns = {'bullish': [], 'bearish': []}
for candle in candle_names:
df[candle] = getattr(talib, candle)(df['open'], df['high'], df['low'],
df['close'])
if df.iloc[-1][candle] > 0:
patterns['bullish'].append(candle)
elif df.iloc[-1][candle] < 0:
patterns['bearish'].append(candle)
df = detect_fractals(df)
if df.iloc[-3]['fractal'] > 0:
patterns['bullish'].append('fractal')
elif df.iloc[-3]['fractal'] < 0:
patterns['bearish'].append('fractal')
#print(df.iloc[-1])
return patterns
def Top_Pattern_recognition(dataframes, n_patterns, rep_ind):
candle_names = ta.get_function_groups()['Pattern Recognition']
candle_names
df = dataframes.copy()
# extract OHLC
op = df['Open']
hi = df['High']
lo = df['Low']
cl = df['Close']
# create columns for each pattern
l = dataframes.columns
cols = [
i for i in l
if i != 'Open' and i != 'Close' and i != 'High' and i != 'Low'
]
df.drop(cols, axis=1, inplace=True)
for candle in candle_names:
# below is same as;
# df["CDL3LINESTRIKE"] = talib.CDL3LINESTRIKE(op, hi, lo, cl)
df[candle] = getattr(ta, candle)(op, hi, lo, cl)
# =============================================================================
# Ranking the patterns
# We successfully extracted candlestick patterns using TA-Lib. With few lines of code, we can condense
# this sparse information into a single column with pattern labels. But first, we need to handle the
# cases where multiple patterns are found for a given candle. To do that, we need a performance metric
# to compare patterns. We will use the “Overall performance rank” from the patternsite.
# =============================================================================
cols = [
i for i in df.columns
if i != 'Open' and i != 'Close' and i != 'High' and i != 'Low'
]
df_abs = df.copy()
df_ = df_abs.abs()
df_["sum"] = df_.sum(axis=1)
df['sum'] = df_['sum']
d = df[df['sum'] > 100 * rep_ind]
d1 = d.sort_values(by='sum', ascending=False)[n_patterns:]
d1 = d1.drop(d1.columns[d1.eq(0).all()], axis=1)
return d1
def enrichTalibValues(inputDict, features, result) :
headers=''
for aType in features:
for func in talib.get_function_groups()[aType]:
pf = abstract.Function(func)
try:
pf.input_arrays = inputDict
pf.run()
outputs = pf.outputs
if isinstance(outputs, list):
for anItem , name in zip(outputs, pf.output_names):
result = numpy.hstack((result, anItem.reshape(anItem.size,1)))
headers = ','.join([headers, name])
else:
result = numpy.hstack((result, outputs.reshape(outputs.size,1)))
headers = ','.join([headers,func])
except:
print
return result, headers
def onTecFunctionSearched(self, function_ui):
text = self.search_tec_function.text().strip().lower()
self.tec_tree.clear()
self.tec_tree = function_ui.findChild(QTreeWidget, "tec_tree")
if text != "":
talib_functions = talib.get_functions()
for i in talib_functions:
if text in i.lower():
node = QTreeWidgetItem(self.tec_tree)
node.setText(0, i)
else:
talib_groups = talib.get_function_groups()
for key in talib_groups.keys():
node = QTreeWidgetItem(self.tec_tree)
node.setText(0, key)
for value in talib_groups[key]:
sub_node = QTreeWidgetItem(node)
sub_node.setText(0, value)
def get_group_api():
# add signal lines to api
mdic = dict(signal_line=dict(fun_name='signal_line',
params=dict(value=0),
settings=dict(shift=0),
outputs=dict(real=['line', 'solid'])))
# add ohlcv to api
for field_name in fields:
mdic[field_name] = dict(fun_name=field_name.lower(),
settings=dict(shift=0),
outputs=dict(real=['line', 'solid']))
mdic['Volume']['outputs']['real'][0] = 'column'
# add talib functions
dic = dict()
# keys are group names
# value are indicators in a group
for key, value in talib.get_function_groups().items():
key = Rename_group_name_dic[key]
if key:
idic = dict()
for fun_name in value:
if fun_name not in Bad_indicators:
name = fun_name + '(' + talib.abstract.Function(
fun_name).info['display_name'] + ')'
idic[name] = dict(fun_name=fun_name,
settings=dict(shift=0))
idic[name]['params'] = get_parameter(fun_name)
output_dic = dict()
for li in list(
talib.abstract.Function(
fun_name).output_flags.items()):
output_dic[li[0]] = ['line', 'solid']
if li[1][0] == 'Histogram':
output_dic[li[0]][0] = 'column'
elif li[1][0] == 'Dashed Line':
output_dic[li[0]][1] = 'Dash'
elif li[1][0] == 'star':
output_dic[li[0]][1] = 'Dot'
idic[name]['outputs'] = output_dic
dic[key] = idic
dic['Historical Data'] = mdic
return dic
def get_all_candlesticks(df):
# get all candlestick pattern functions
# all functions take open, high, low, close, except
# abandoned baby, dark cloud cover, evening doji star, evening star, mat hold, morning doji star, morning star, which all have
# penetration=0
# penetration is percentage of penetration of one candle into another, e.g. 0.3 is 30%
o, h, l, c = get_ohlc_for_talib(df)
pattern_functions = talib.get_function_groups()['Pattern Recognition']
# for dataframe; cuts out CDL, so will make things like 'CDLHARAMI' just 'HARAMI'
column_names = [p[3:] for p in pattern_functions]
# for bearish, seems to give -100, for nothing, 0, for bullish, 100
data = {}
for col, p in zip(column_names, pattern_functions):
# detect candlestick pattern
data[col] = getattr(talib, p)(o, h, l, c)
# would get 3-week exponential moving average of pattern
# data[col + '_mva_15'] = talib.EMA(data[col], timeperiod=15)
cs_df = pd.DataFrame(data)
cs_df.index = df.index
full_df = pd.concat([df, cs_df], axis=1)
return cs_df, full_df
def getPatternIndicators(df):
high = df['High']
low = df['Low']
close = df['Close']
open = df['Open']
volume = df['Volume']
'''
df['2CROWS'] = ta.CDL2CROWS(open, high, low, close)
df['3BLACKCROWS'] = ta.CDL3BLACKCROWS(open, high, low, close)
df['3INSIDE'] = ta.CDL3INSIDE(open, high, low, close)
df['3LINESTRIKE'] = ta.CDL3LINESTRIKE(open, high, low, close)
df['3OUTSIDE'] = ta.CDL3OUTSIDE(open, high, low, close)
df['3STARSOUTH'] = ta.CDL3STARSINSOUTH(open, high, low, close)
df['3WHITESOLDIERS'] = ta.CDL3WHITESOLDIERS(open, high, low, close)
df['ABANDONEDBABY'] = ta.CDLABANDONEDBABY(open, high, low, close, penetration=0)
df['ADVANCEBLOCK'] = ta.CDLADVANCEBLOCK(open, high, low, close)
df['BELTHOLD'] = ta.CDLBELTHOLD(open, high, low, close)
df['BREAKAWAY'] = ta.CDLBREAKAWAY(open, high, low, close)
'''
group = ta.get_function_groups()
for i in group['Pattern Recognition']:
print i
method = getattr(ta, i)
df[i] = method(open, high, low, close)
def get_groups_of_functions(self):
ret = talib.get_function_groups()
ret.pop('Math Operators')
ret.pop('Math Transform')
ret.pop('Statistic Functions')
return ret
def get_groups_markdown():
"""Generate markdown for function groups using the Abstract API
Returns a dictionary of group_name -> documentation for group functions
"""
def unpluralize(noun):
if noun.endswith('s'):
if len(noun) > 2 and noun[-2] not in ["'", 'e']:
return noun[:-1]
return noun
doc_links = get_doc_links()
ret = {}
for group, funcs in talib.get_function_groups().items():
h1 = '# %s' % unpluralize(group)
h1 = h1 + ' Functions' if 'Function' not in h1 else h1 + 's'
group_docs = [h1]
for func in funcs:
# figure out this function's options
f = Function(func)
inputs = f.info['input_names']
if 'price' in inputs and 'prices' in inputs:
names = [inputs['price']]
names.extend(inputs['prices'])
input_names = ', '.join(names)
elif 'prices' in inputs:
input_names = ', '.join(inputs['prices'])
else:
input_names = ', '.join([x for x in inputs.values() if x])
params = ', '.join(
['%s=%i' % (param, default)
for param, default in f.info['parameters'].items()])
outputs = ', '.join(f.info['output_names'])
# print the header
group_docs.append('### %s - %s' % (func, f.info['display_name']))
if f.function_flags and 'Function has an unstable period' in f.function_flags:
group_docs.append('NOTE: The ``%s`` function has an unstable period. ' % func)
# print the code definition block
group_docs.append("```python")
if params:
group_docs.append('%s = %s(%s, %s)' % (
outputs, func.upper(), input_names, params))
else:
group_docs.append('%s = %s(%s)' % (
outputs, func.upper(), input_names))
group_docs.append("```\n")
# print extra info if we can
if func in doc_links:
group_docs.append('Learn more about the %s at [tadoc.org](%s). ' % (
f.info['display_name'], doc_links[func]))
group_docs.append('\n[Documentation Index](../doc_index.html)')
group_docs.append('[FLOAT_RIGHTAll Function Groups](../funcs.html)')
ret[slugify(group)] = '\n'.join(group_docs) + '\n'
return ret
"""
indicators.py
uses talib to add indicator data to market data
"""
import talib # @UnresolvedImport
FUNCS = talib.get_functions()
FUNC_GROUPS = talib.get_function_groups()
from utils.logger import fx_logger
log = fx_logger(__name__)
log.setLevel("DEBUG")
class Calculator(object):
"""
wrapper to allow calling of indicator functions
(just ta-lib functions for now)
"""
def __call__(self, name, *data, **kwargs):
if name in FUNCS:
meth = getattr(talib, name)
return meth(*data, **kwargs)
else:
log.error("talib does not provide {0}".format(name))
return
def supported():
return ta.get_function_groups()
_STOCK_DATA_ADDED_COLUMNS = EnumDict(open_trigger='OPEN_TRIGGER', close_trigger='CLOSE_TRIGGER')
STOCK_DATA_COLUMNS = EnumDict(date='Date') + RAW_PARAMETERS + _STOCK_DATA_ADDED_COLUMNS + SUPPORTED_INDICATORS
TECHNICAL_PARAMETER = RAW_PARAMETERS + SUPPORTED_INDICATORS
NUMERIC_VALUE = EnumDict(numeric_value='NUMERIC_VALUE') # for technical param that is pure value (float)
TRADE_DIRECTIONS = EnumDict(long='LONG', short='SHORT')
"""
The next Enum class contains all the order relations between indicators.
"""
RELATIONS = EnumDict(greater='GREATER', less='LESS', crossover='CROSSOVER', crossover_below='CROSSOVER_BELOW',
crossover_above='CROSSOVER_ABOVE')
# TA-Lib technical indicators by groups
INDICATORS_GROUPS = EnumDict(zip([k.lower().replace(' ', '_') for k in talib.get_function_groups().keys()],
[k for k in talib.get_function_groups().keys()]))
CYCLE_INDICATORS = EnumDict(zip([ind.lower() for ind in talib.get_function_groups().get(INDICATORS_GROUPS.cycle_indicators)],
[ind for ind in talib.get_function_groups().get(INDICATORS_GROUPS.cycle_indicators)]))
MOMENTUM_INDICATORS = EnumDict(zip([ind.lower() for ind in talib.get_function_groups().get(INDICATORS_GROUPS.momentum_indicators)],
[ind for ind in talib.get_function_groups().get(INDICATORS_GROUPS.momentum_indicators)]))
OVERLAP_STUDIES_INDICATORS = EnumDict(zip([ind.lower() for ind in talib.get_function_groups().get(INDICATORS_GROUPS.overlap_studies)],
[ind for ind in talib.get_function_groups().get(INDICATORS_GROUPS.overlap_studies)]))
VOLATILITY_INDICATORS = EnumDict(zip([ind.lower() for ind in talib.get_function_groups().get(INDICATORS_GROUPS.volatility_indicators)],
[ind for ind in talib.get_function_groups().get(INDICATORS_GROUPS.volatility_indicators)]))
VOLUME_INDICATORS = EnumDict(zip([ind.lower() for ind in talib.get_function_groups().get(INDICATORS_GROUPS.volume_indicators)],
[ind for ind in talib.get_function_groups().get(INDICATORS_GROUPS.volume_indicators)]))
# PRICE_TRANSFORM = EnumDict(zip([ind.lower() for ind in talib.get_function_groups().get(INDICATORS_GROUPS.price_transform)],
# [ind for ind in talib.get_function_groups().get(INDICATORS_GROUPS.price_transform)]))
# # d/dt (ema3-ema7)
# ematimesd37 = noNaN(np.c_[ times[1:], (ema3[1:]-ema7[:-1])/inputs['close'][1:] ], 1)
# addToDict(ematimesd37, ticker, "EMAD37")
# know sure thing http://stockcharts.com/school/doku.php?id=chart_school:technical_indicators:know_sure_thing_kst
ksttimes = noNaN(np.c_[ times, ind.kst(inputs['close']) ], 1)
addToDict(ksttimes, ticker, "KST")
# all indicators below
indicators = ["BBANDS","DEMA","EMA","SAR","SAREXT", "ADX","ADXR","APO","AROON", \
"AROONOSC","BOP","CCI","CMO","DX","MACD","MACDEXT", "MACDFIX","MFI", \
"MINUS_DI","MINUS_DM","MOM","ROC", "RSI","STOCH","STOCHF","STOCHRSI", \
"WILLR","AD","ADOSC","HT_DCPERIOD","HT_DCPHASE", "HT_SINE","NATR","TRANGE" ]
functiongroups = ta.get_function_groups()
for fname in indicators:
fn = ta.abstract.Function(fname)
# skip candlesticks crap
if(fname in functiongroups['Pattern Recognition']): continue
params = {}
outputs = fn(inputs, **params)
numoutputs = len(fn.info['output_names'])
norm = 1.0
if(fname in functiongroups['Overlap Studies']): norm = inputs['close'] # normalize these indicators via closing price
if(fname in functiongroups['Price Transform']): norm = inputs['close']
def get_func_list():
talib_func = talib.get_function_groups()
return talib_func
bbands = Function('bbands', input_arrays)
bbands.set_function_parameters(timePeriod=20, nbDevUp=2, nbDevDown=2)
upper, middle, lower = bbands() # multiple output values unpacked (these will always have the correct order)
kama = Function('kama').run(input_arrays) # alternative run() calling method.
plot(odata, upper, middle, lower, kama)
def plot(odata, upper, middle, lower, kama):
pylab.plot(r, idata, 'b-', label="original")
pylab.plot(r, odata, 'g-', label="MA")
pylab.plot(r, upper, 'r-', label="Upper")
pylab.plot(r, middle, 'r-', label="Middle")
pylab.plot(r, lower, 'r-', label="Lower")
pylab.plot(r, kama, 'g', label="KAMA")
pylab.legend()
pylab.show()
if __name__ == '__main__':
print 'All functions (sorted by group):'
groups = talib.get_function_groups()
for group, functions in sorted(groups.items()):
print '%s functions: ' % group, functions
if len(sys.argv) == 1 or sys.argv[1] == 'func':
print 'Using talib.func'
func_example()
else:
print 'Using talib.abstract'
abstract_example()
