def useCurrentCSV(self):
self.dataFile = pullData.read_csv('{0} ohlc.csv'.format(self.symbol),
index_col='Date',
parse_dates=True)
#print('self.dataFile print:', self.dataFile[2:3])
print("Entered stxSetFile1b.py to create new file")
return self.dataFile
def createCSV(self):
print
self.timeSeries0.to_csv('{0} ohlc.csv'.format(self.symbol))
self.dataFile = pullData.read_csv('{0} ohlc.csv'.format(self.symbol),
index_col='Date',
parse_dates=True)
return self.dataFile
def createCSV(self):
print
# self.timeSeries0.to_csv('{0} ohlc.csv'.format(self.symbol))
self.timeSeries0.to_csv('../{0} ohlc.csv'.format(self.symbol))
# self.dataFile = pullData.read_csv('{0} ohlc.csv'.format(self.symbol), index_col='Date',parse_dates=True)
self.dataFile = pullData.read_csv('../{0} ohlc.csv'.format(self.symbol), index_col='Date', parse_dates=True)
return self.dataFile
def createCSV(self):
# root = Tk()
# # dateStart = input("Enter start date (format yyyymmdd): ")
# # dateEnd = input("Enter end date (format yyyymmdd): ")
# root.mainloop()
self.timeSeries0.to_csv('{0} ohlc.csv'.format(self.symbol))
self.dataFile = pullData.read_csv('{0} ohlc.csv'.format(self.symbol), index_col='Date',parse_dates=True)
return self.dataFile
def createCSV(self):
# root = Tk()
# # dateStart = input("Enter start date (format yyyymmdd): ")
# # dateEnd = input("Enter end date (format yyyymmdd): ")
# root.mainloop()
self.timeSeries0.to_csv('{0} ohlc.csv'.format(self.symbol))
self.dataFile = pullData.read_csv('{0} ohlc.csv'.format(self.symbol),
index_col='Date',
parse_dates=True)
return self.dataFile
def useCurrentCSV(self):
# self.dataFile = pullData.read_csv('{0} ohlc.csv'.format(self.symbol), index_col='Date',parse_dates=True)
self.dataFile = pullData.read_csv('../{0} ohlc.csv'.format(self.symbol), index_col='Date', parse_dates=True)
#print('self.dataFile print:', self.dataFile[2:3])
print("Entered stxSetFile1b.py to create new file")
return self.dataFile
# setup zmq push
CONTEXT = zmq.CONTEXT()
TWITTER_PUB_ADDR = 'tcp://127.0.0.1:6050'
FR_RCV_ADDR = 'tcp://127.0.0.1:6051'
# twitter access codes
# need these in real life :)
CONSUMER_KEY = ''
CONSUMER_SECRET = ''
ACCESS_TOKEN = ''
ACCESS_TOKEN_SECRET = ''
TWITTER_SOURCES_FILE = "nasdaq_finance_400_2014_05_20.csv"
TWITTER_SOURCES = read_csv(TWITTER_SOURCES_FILE)
def gen_follow_symbols():
""" Create a list of symbols to follow from the csv file """
follow_syms = []
for i, ticker in enumerate(TWITTER_SOURCES.Symbol.values):
ticker = "$" + str(ticker)
follow_syms.append(ticker)
return follow_syms
def get_ticker_info(data):
""" Lookup symbol's sector in Nasdaq """
m_class = TWITTER_SOURCES[
TWITTER_SOURCES['Symbol'].str.contains(
import pandas.io.data as pd
import numpy as np
from sklearn import svm, linear_model
import decisiontree_lzl
import naivebayes_lc
train_data = np.array(pd.read_csv("train_data.csv").values.tolist(), dtype=np.int_)
test_data = np.array(pd.read_csv("test_data.csv").values.tolist(), dtype=np.int_)
train_data_b = np.array(pd.read_csv("train_data_b.csv").values.tolist(), dtype=np.int_)
test_data_b = np.array(pd.read_csv("test_data_b.csv").values.tolist(), dtype=np.int_)
train_X = train_data[:,:10]
train_Y = train_data[:,10]
test_X = test_data[:,:10]
test_Y = test_data[:,10]
train_X_b = train_data_b[:,:10]
train_Y_b = train_data_b[:,10]
test_X_b = test_data_b[:,:10]
test_Y_b = test_data_b[:,10]
clf_svm = svm.SVC()
clf_svm.fit(train_X, train_Y)
svm_score = clf_svm.score(test_X, test_Y)
clf_svm.fit(train_X_b, train_Y_b)
svm_score_b = clf_svm.score(test_X_b, test_Y_b)
clf_lr = linear_model.LogisticRegression()
clf_lr.fit(train_X, train_Y)
lr_score = clf_lr.score(test_X, test_Y)
def iterlines(x, window, fromdate='1900-01-01', todate=None, charts=True,
log_scale=False, directory=None):
"""
Turn minitrends to iterative process more easily adaptable to
implementation in simple trading systems; allows backtesting functionality.
x - ticker symbol or data set
window - float defining how far back the algorithm checks for critical
values
fromate - start date for the stock data
todate - end date for the stock data
charts - boolean value saying whether to print chart to screen
log_scale - converts imported data to logarithmic scale
directory - directory in which data may be found to save on import speed
"""
# Import packages
import pandas.io.data as pd
import numpy as np
from matplotlib.pyplot import subplot, plot, show, title, grid
# Check inputs and get data
if type(x) == str:
if directory is None:
if todate is None:
y = pd.DataReader(x, 'yahoo', fromdate)
y = np.array(y['Adj Close'])
else:
y = pd.DataReader(x, 'yahoo', fromdate, todate)
y = np.array(y['Adj Close'])
else:
y = pd.read_csv(directory + x + '.csv')
if (fromdate == '1900-01-01') & (todate is None):
y = np.array(y['Adj Close'])
elif (fromdate == '1900-01-01') & (todate is not None):
todate = np.where(y.Date == str(todate)[0:10])[0][0]
y = np.array(y['Adj Close'])[0:todate]
elif (fromdate != '1900-01-01') & (todate is None):
fromdate = np.where(y.Date == str(fromdate)[0:10])[0][0]
y = y['Adj Close'][fromdate:]
elif (fromdate != '1900-01-01') & (todate is not None):
fromdate = np.where(y.Date == str(fromdate)[0:10])[0][0]
todate = np.where(y.Date == str(todate)[0:10])[0][0]
y = y['Adj Close'][fromdate:todate]
else:
y = x
# Turn to log scale if desired
if log_scale is True:
y = np.log(y)
# Change to log scale if desired
if log_scale is True:
y = np.log(y)
if window < 1:
window = int(window * len(y))
x = np.arange(0, len(y))
xmax = np.array(x[0])
xmin = np.array(x[0])
ymax = np.array(y[0])
ymin = np.array(y[0])
for i in x[window:]:
if y[i] > max(y[i-window:i]):
ymax = np.append(ymax, y[i])
xmax = np.append(xmax, x[i])
if y[i] < min(y[i-window:i]):
ymin = np.append(ymin, y[i])
xmin = np.append(xmin, x[i])
# Plot results if desired
if charts is True:
plot(x, y)
plot(xmax, ymax, 'o')
plot(xmin, ymin, 'o')
grid(True)
show()
return R[-1]
def segtrend(x, segments=2.0, charts=True, fromdate='1900-01-01', todate=None,
log_scale=False, directory=None):
"""
Trendline algorithm that segments data into pieces and finds trendlines
using those subsets.
x - ticker symbol or data set
threshold - ratio from 0 to 1 giving the desired extrema barrier window
charts - boolean, whether or not to print charts to screen
fromdate - when to start pulling stock data (defaults to all data)
todate - when to stop pulling stock data (if none, defaults to most recent)
log_scale - converts data to logarithmic scale
directory - directory in which data may be found to save on import speed
"""
# IMPORT PACKAGES
import pandas.io.data as pd
from matplotlib.pyplot import plot, grid, show, title
import numpy as np
# Check inputs and get data
if type(x) == str:
if directory is None:
if todate is None:
y = pd.DataReader(x, 'yahoo', fromdate)
y = np.array(y['Adj Close'])
else:
y = pd.DataReader(x, 'yahoo', fromdate, todate)
y = np.array(y['Adj Close'])
else:
y = pd.read_csv(directory + x + '.csv')
if (fromdate == '1900-01-01') & (todate is None):
y = np.array(y['Adj Close'])
elif (fromdate == '1900-01-01') & (todate is not None):
todate = np.where(y.Date == str(todate)[0:10])[0][0]
y = np.array(y['Adj Close'])[0:todate]
elif (fromdate != '1900-01-01') & (todate is None):
fromdate = np.where(y.Date == str(fromdate)[0:10])[0][0]
y = y['Adj Close'][fromdate:]
elif (fromdate != '1900-01-01') & (todate is not None):
fromdate = np.where(y.Date == str(fromdate)[0:10])[0][0]
todate = np.where(y.Date == str(todate)[0:10])[0][0]
y = y['Adj Close'][fromdate:todate]
else:
y = x
if log_scale:
y = np.log(y) # change to log scale if desired
# Implement trendlines
segments = int(segments)
maxima = np.ones(segments)
minima = np.ones(segments)
segsize = int(len(y)/(segments))
for i in range(1, segments+1):
ind2 = i*segsize
ind1 = ind2 - segsize
maxima[i-1] = max(y[ind1:ind2])
minima[i-1] = min(y[ind1:ind2])
# Find the indexes of these maxima in the data
x_maxima = np.ones(segments)
x_minima = np.ones(segments)
for i in range(0, segments):
x_maxima[i] = np.where(y == maxima[i])[0][0]
x_minima[i] = np.where(y == minima[i])[0][0]
# Return some output
if charts:
plot(y)
for i in range(0, segments-1):
maxslope = (maxima[i+1] - maxima[i]) / (x_maxima[i+1] - x_maxima[i])
a_max = maxima[i] - (maxslope * x_maxima[i])
b_max = maxima[i] + (maxslope * (len(y) - x_maxima[i]))
maxline = np.linspace(a_max, b_max, len(y))
minslope = (minima[i+1] - minima[i]) / (x_minima[i+1] - x_minima[i])
a_min = minima[i] - (minslope * x_minima[i])
b_min = minima[i] + (minslope * (len(y) - x_minima[i]))
minline = np.linspace(a_min, b_min, len(y))
if charts:
plot(maxline, 'g')
plot(minline, 'r')
# OUTPUT
grid(True)
show()
return x_maxima, maxima, x_minima, minima
# Run it
# segtrend(x='goog',
segments=5,
charts=True,
fromdate='1900-01-01',
todate=None,
directory='/Users/JAmos/Dropbox/Research/Trading/Data Dump/')