def mrTest(filename):
stocks = ['C', 'FDX', 'KO', 'MSFT', 'SBUX', 'NFLX', 'LUV']
for s in stocks:
csvFile = open(s+filename, 'a')
f = csv.writer(csvFile)
table = {}
f.writerow([s])
f.writerow(['t', 'R^2', 'AIC'])
path = createPath(s + '/' + s + 'mrValidation.csv')
models = loadModels(path, "mr")
remainingStocks = copy.deepcopy(stocks)
remainingStocks.remove(s)
for line in models:
for rs in remainingStocks:
i = line['timewindow']
featurePath = createPath(rs + "/")
features = loadFeatures(featurePath, i)
prices = []
ema = []
rsi = []
macd = []
for row in features:
prices.append(row['price'])
ema.append(row['ema'])
rsi.append(row['rsi'])
macd.append(row['macd'])
y = np.array(prices)
x = np.vstack([ema, rsi, macd]).T
validationStats = anova.ols(y, x, [line['constant'], line['ema'], line['rsi'], line['macd']])
r2 = validationStats.R2
aic = validationStats.ll()[1]
# average out the results over k folds
stats = [r2, aic]
stats = [val/float(len(remainingStocks)) for val in stats]
if i in table:
table[i] = [sum(item) for item in izip(table[i], stats)]
else:
table[i] = stats
print i
print stats
# write the averaged results
for i in range(2, 202, 2):
table[i].insert(0, i)
print table[i]
f.writerow(table[i])
csvFile.flush()
def crossValidationFeatures(featureLocation, filename, algorithm, k=10):
csvFile = open(filename, 'a')
f = csv.writer(csvFile)
f.writerow(['t', 'constant', 'EMA', 'RSI' , 'MACD', 'R^2', 'AIC'])
table = {}
for i in range(2, 202, 2):
features = loadFeatures(featureLocation, i)
bestSol = []
bestR2 = 0
for training, validation in k_fold_cross_validation(features, k):
algorithm.setFeatures(training)
sol = algorithm.findSol(i)
prices = []
ema = []
rsi = []
macd = []
for row in validation:
prices.append(row['price'])
ema.append(row['ema'])
rsi.append(row['rsi'])
macd.append(row['macd'])
y = np.array(prices)
x = np.vstack([ema, rsi, macd]).T
validationStats = anova.ols(y, x, [sol['constant'], sol['EMA'], sol['RSI'], sol['MACD']])
r2 = validationStats.R2
aic = validationStats.ll()[1]
if (r2 > bestR2):
bestR2 = r2
bestSol = [sol['constant'], sol['EMA'], sol['RSI'], sol['MACD']]
# average out the results over k folds
stats = [r2, aic]
stats = [val/float(k) for val in stats]
if i in table:
table[i] = [sum(item) for item in izip(table[i], stats)]
else:
table[i] = stats
for j in range(len(bestSol)):
table[i].insert(j, bestSol[j])
print i
# write the averaged results
table[i].insert(0, i)
print table[i]
f.writerow(table[i])
csvFile.flush()
def emaTest(filename):
stocks = ['C', 'FDX', 'KO', 'MSFT', 'SBUX', 'NFLX', 'LUV']
for s in stocks:
csvFile = open(s + filename, 'a')
f = csv.writer(csvFile)
table = {}
f.writerow([s])
f.writerow(['t', 'R^2', 'AIC'])
path = createPath(s + '/' + s + 'emaValidation.csv')
models = loadModels(path, "emaOnly")
remainingStocks = copy.deepcopy(stocks)
remainingStocks.remove(s)
for line in models:
for rs in remainingStocks:
i = line['timewindow']
featurePath = createPath(rs + "/")
features = loadFeatures(featurePath, i)
prices = []
ema = []
for row in features:
prices.append(row['price'])
ema.append(row['ema'])
y = np.array(prices)
x = np.vstack([ema]).T
validationStats = anova.ols(y, x,
[line['constant'], line['ema']])
r2 = validationStats.R2
aic = validationStats.ll()[1]
# average out the results over k folds
stats = [r2, aic]
stats = [val / float(len(remainingStocks)) for val in stats]
if i in table:
table[i] = [sum(item) for item in izip(table[i], stats)]
else:
table[i] = stats
print i
print stats
# write the averaged results
for i in range(2, 202, 2):
table[i].insert(0, i)
print table[i]
f.writerow(table[i])
csvFile.flush()
def simpleCrossValidationFeatures(featureLocation, filename, algorithm, k=10):
csvFile = open(filename, 'a')
f = csv.writer(csvFile)
f.writerow(['t', 'constant', 'EMA', 'R^2', 'AIC'])
table = {}
for i in range(2, 202, 2):
features = loadFeatures(featureLocation, i)
bestSol = []
bestR2 = 0
for training, validation in k_fold_cross_validation(features, k):
algorithm.setFeatures(training)
sol = algorithm.findSol(i)
prices = []
ema = []
for row in validation:
prices.append(row['price'])
ema.append(row['ema'])
y = np.array(prices)
x = np.vstack([ema]).T
validationStats = anova.ols(y, x, [sol['constant'], sol['EMA']])
r2 = validationStats.R2
aic = validationStats.ll()[1]
if (r2 > bestR2):
bestR2 = r2
bestSol = [sol['constant'], sol['EMA']]
# average out the results over k folds
stats = [r2, aic]
stats = [val / float(k) for val in stats]
if i in table:
table[i] = [sum(item) for item in izip(table[i], stats)]
else:
table[i] = stats
print i
print stats
for j in range(len(bestSol)):
table[i].insert(j, bestSol[j])
# write the averaged results
for i in range(2, 202, 2):
table[i].insert(0, i)
print table[i]
f.writerow(table[i])
csvFile.flush()
def crossValidation(filename, timeseries, algorithm, k=10):
csvFile = open(filename, 'a')
f = csv.writer(csvFile)
f.writerow(['t', 'constant', 'EMA', 'RSI', 'MACD', 'R^2', 'AIC'])
timeseries = timeseries.items()
table = {}
for training, validation in k_fold_cross_validation(timeseries, k):
trainingDict = OrderedDict()
for key, value in training:
trainingDict[key] = value
validationDict = OrderedDict()
for key, value in validation:
validationDict[key] = value
tr = tf.TechnicalFeatures(trainingDict)
vd = tf.TechnicalFeatures(validationDict)
algorithm.setFeatures(tr)
for i in range(2, 202, 2):
sol = algorithm.findSol(i)
prices, ema, rsi, macd = vd.getTimewindow(i)
y = np.array(prices)
x = np.vstack([ema, rsi, macd]).T
validationStats = anova.ols(
y, x, [sol['constant'], sol['EMA'], sol['RSI'], sol['MACD']])
r2 = validationStats.R2
aic = validationStats.ll()[1]
# average out the results over k folds
stats = [
sol['constant'], sol['EMA'], sol['RSI'], sol['MACD'], r2, aic
]
stats = [val / float(k) for val in stats]
if i in table:
table[i] = [sum(item) for item in izip(table[i], stats)]
else:
table[i] = stats
print i
print stats
# write the averaged results
for i in range(2, 202, 2):
table[i].insert(0, i)
print table[i]
f.writerow(table[i])
csvFile.flush()
def simpleCrossValidation(filename, timeseries, algorithm, k=10):
csvFile = open(filename, 'a')
f = csv.writer(csvFile)
f.writerow(['t', 'constant', 'EMA', 'R^2', 'AIC'])
timeseries = timeseries.items()
table = {}
for training, validation in k_fold_cross_validation(timeseries, k):
trainingDict = OrderedDict()
for key, value in training:
trainingDict[key] = value
validationDict = OrderedDict()
for key, value in validation:
validationDict[key] = value
tr = tf.TechnicalFeatures(trainingDict)
vd = tf.TechnicalFeatures(validationDict)
algorithm.setFeatures(tr)
for i in range(2, 202, 2):
sol = algorithm.findSol(i)
prices, ema, rsi, macd = vd.getTimewindow(i)
y = np.array(prices)
x = np.vstack([ema]).T
validationStats = anova.ols(y, x, [sol['constant'], sol['EMA']])
r2 = validationStats.R2
aic = validationStats.ll()[1]
# average out the results over k folds
stats = [sol['constant'], sol['EMA'], r2, aic]
stats = [val/float(k) for val in stats]
if i in table:
table[i] = [sum(item) for item in izip(table[i], stats)]
else:
table[i] = stats
print i
print stats
# write the averaged results
for i in range(2, 202, 2):
table[i].insert(0, i)
print table[i]
f.writerow(table[i])
csvFile.flush()