def Enviro(stockList, Investment, numSpecies, genCount, shareCount, trainStart,
trainEnd, testStart, testEnd, SaveNew, geneTest, evolutionPlot):
avg = []
mx = []
graph = {}
g = []
TOPV = 0
TOP = 0
qg = []
Environment = Strategy(stockList)
pInvest = Investment
startDay = trainStart
DAYS = trainEnd
inputList = {}
startPriceList = {}
endPriceList = {}
#Make Portfolios
for individual in range(numSpecies):
Environment.addPortfolio("p" + str(individual), pInvest)
#Set inputList
try:
inputList = pickle.load(open("inputListA.p", "rb"))
startPriceList = pickle.load(open("startPriceListA.p", "rb"))
endPriceList = pickle.load(open("endPriceListA.p", "rb"))
except (IOError):
print "Writing New Pickle"
for d in range(startDay, DAYS):
if d in inputList:
pass
else:
print "Adding Day", d
Environment.setInputs("LUVBBF3", d, d + 1)
inputList[d] = (Environment.getInputsF().copy())
startPriceList[d] = Environment.getRecentPrice()
endPriceList[d] = Environment.getEndPrice()
pickle.dump(inputList, open("inputListA.p", "wb"))
pickle.dump(startPriceList, open("startPriceListA.p", "wb"))
pickle.dump(endPriceList, open("endPriceListA.p", "wb"))
#Make lists for randomization of initial population
inputs = inputList[startDay]
variables = inputs.keys()
right = [[], []]
weight = []
fitness = {}
delimeters = ["<", ">"]
for valTop in np.arange(-1, 1.1, .1):
right[0].append(round(valTop, 2))
for y in variables:
right[1].append(y)
for z in np.arange(0, 1, .01):
weight.append(round(z, 3))
#Construct initial population
for portfolio in Environment.getPortfolios().keys():
p = Environment.getPortfolios()[portfolio]
p.addSpecies(inputs)
for variable in Environment.getPortfolios()[portfolio].getSpecies(
).getGenes().keys():
a = random.randint(0, 1)
b = random.randint(0, len(right[a]) - 1)
c = delimeters[random.randint(0, 1)]
d = weight[random.randint(0, 99)]
Environment.getPortfolios()[portfolio].getSpecies().addGene(
variable, right[a][b], c, d)
Environment.getPortfolios()[portfolio].getSpecies().setThresh(
random.randrange(
-len(Environment.getPortfolios()
[portfolio].getSpecies().getGenes()), 1) + 0.0,
random.randrange(
0,
len(Environment.getPortfolios()
[portfolio].getSpecies().getGenes()) + 1) + 0.0)
Environment.getPortfolios()[portfolio].getSpecies().setShareFactor(
(random.randrange(1, 100)))
# Start loop of days and generations
manager = mp.Manager()
portDict = manager.dict()
for gen in range(genCount):
portDict.clear()
t1 = time.time()
print "Generation : ", gen + 1
pList = []
for portfolio in Environment.getPortfolios().keys():
proc = mp.Process(target=MultiProcess1,
args=(portDict, Environment, startDay, DAYS,
inputList, startPriceList, endPriceList,
stockList, portfolio))
pList.append(proc)
proc.start()
for proc in pList:
if proc.is_alive():
proc.join()
for port in portDict.keys():
Environment.getPortfolios()[port] = portDict[port]
#Get fitness of species
for portfolio in Environment.getPortfolios().keys():
#fitness[portfolio]=(Environment.getPortfolios()[portfolio].getCorrectList().count(1))*1.0/len(Environment.getPortfolios()[portfolio].getCorrectList())
fitness[portfolio] = (Environment.getPortfolios(
)[portfolio].getCorrectList().count(1)) * 1.0 / len(
Environment.getPortfolios()[portfolio].getCorrectList())
# Environment.getPortfolios()[portfolio].balanceList[-1]
#Find top 50%
sorted_fitness = sorted(fitness.items(), key=operator.itemgetter(1))
if (sorted_fitness[-1][-1] > TOPV):
TOP = sorted_fitness[-1][0]
TOPV = sorted_fitness[-1][1]
#Get bottom 50%
bottom = sorted_fitness[0:(numSpecies / 2)]
#Choose 50 pairs of top 50
top = sorted_fitness[numSpecies / 2:numSpecies]
topPair = []
# for valTop in range(len(top)):
# pair=valTop
# while top[pair][0]==top[valTop][0]:
# pair=random.randint(0,len(top)-1)
# topPair.append((top[valTop][0],top[pair][0]))
pairs = []
for valTop in range(len(top)):
for val2 in range(valTop + 1, len(top)):
pairs.append((top[valTop][0], top[val2][0]))
np.random.shuffle(pairs)
topPair = pairs[:numSpecies]
#For portfolio in 50 worst, set it equal to the breeding of a pair of top 50
for valBot in range(len(bottom)):
a = Environment.getPortfolios()[topPair[valBot]
[0]].getSpecies().copy()
b = Environment.getPortfolios()[topPair[valBot]
[1]].getSpecies().copy()
Environment.getPortfolios()[bottom[valBot][0]].resetAll(pInvest)
Environment.getPortfolios()[bottom[valBot][0]].updateSpecies(
a.breed(b).copy())
q = []
for portfolio in range(len(Environment.getPortfolios().keys())):
for port2 in range(portfolio + 1,
len(Environment.getPortfolios().keys())):
if (portfolio != port2):
q.append(
Environment.getPortfolios()[Environment.getPortfolios(
).keys()[portfolio]].getSpecies().numEquals(
Environment.getPortfolios()[
Environment.getPortfolios().keys()
[port2]].getSpecies()))
print np.round(
100.0 * average(q) /
len(Environment.getPortfolios()[Environment.getPortfolios().keys()
[0]].getSpecies().getGenes()),
2), "%"
qg.append(100.0 * average(q) / len(Environment.getPortfolios()[
Environment.getPortfolios().keys()[0]].getSpecies().getGenes()))
print np.round(time.time() - t1, 2), "seconds"
print
print
if (gen != genCount - 1):
for valTop in range(len(top)):
Environment.getPortfolios()[top[valTop][0]].reset(pInvest)
print TOPV, "Fitness"
if (len(Environment.getPortfolios()[TOP].getCorrectList()) == 0):
print "bad things happend there was nothing there, uh hello"
else:
print "TRAIN CORRECT : ", (
Environment.getPortfolios()[TOP].getCorrectList().count(1) *
1.0) / len(Environment.getPortfolios()[TOP].getCorrectList())
print "TRAIN INCORRECT : ", (
Environment.getPortfolios()[TOP].getCorrectList().count(-1) *
1.0) / len(Environment.getPortfolios()[TOP].getCorrectList())
print "TRAIN HELD : ", (
Environment.getPortfolios()[TOP].getCorrectList().count(0) *
1.0) / len(Environment.getPortfolios()[TOP].getCorrectList())
print "TRAIN GUESSED : ", (
Environment.getPortfolios()[TOP].getCorrectList().count(1) +
Environment.getPortfolios()[TOP].getCorrectList().count(-1) *
1.0) / len(Environment.getPortfolios()[TOP].getCorrectList())
print "TRAIN BALANCE : ", Environment.getPortfolios(
)[TOP].getBalanceList()[-1]
pToO = []
for day in range(startDay, DAYS):
inputs = inputList[day]
startPrice = startPriceList[day]
endPrice = endPriceList[day]
# Get boolean statements for the day
for gene in Environment.getPortfolios()[TOP].getSpecies().getGenes(
).keys():
Environment.getPortfolios()[TOP].getSpecies().getGenes(
)[gene].makeBoolean(inputs)
# Get species output and add corresponding behavior
# print Environment.getPortfolios()[portfolio].getBalance()/(0.0+startPrice),Environment.getPortfolios()[portfolio].getShareFactor()
Environment.getPortfolios()[TOP].addBehavior(
Environment.getPortfolios()[TOP].getSpecies().getOutput() >
Environment.getPortfolios()[TOP].getSpecies().getRightThresh(),
("Buy", "LUV",
int(Environment.getPortfolios()[TOP].getShareFactor() *
(Environment.getPortfolios()[TOP].getBalance() /
(0.0 + startPrice))), startPrice))
Environment.getPortfolios()[TOP].addBehavior(
Environment.getPortfolios()[TOP].getSpecies().getOutput() <
Environment.getPortfolios()[TOP].getSpecies().getLeftThresh(),
("Short", "LUV",
int(Environment.getPortfolios()[TOP].getShareFactor() *
(Environment.getPortfolios()[TOP].getBalance() /
(0.0 + startPrice))), startPrice))
Environment.getPortfolios()[TOP].addGeneCorrect(endPrice - startPrice)
#Go through with actions, sellBack at end of day and add balance to balanceList
Environment.getPortfolios()[TOP].makeActions(startPrice)
Environment.getPortfolios()[TOP].sellBack(stockList, endPrice)
Environment.getPortfolios()[TOP].addBalance(
Environment.getPortfolios()[TOP].balance)
pToO.append((Environment.getPortfolios()[TOP].getSpecies().getOutput(),
(Environment.getPortfolios()[TOP].balanceList[-1] -
Environment.getPortfolios()[TOP].balanceList[-2]) / abs(
(Environment.getPortfolios()[TOP].balanceList[-1] -
Environment.getPortfolios()[TOP].balanceList[-2]))))
np.savetxt("pCorrectToOutput.csv", pToO, delimiter=",")
# if evolutionPlot == True:
# plt.plot(Environment.getPortfolios()[TOP].getBalanceList())
# plt.show()
if geneTest == True:
for gene in Environment.getPortfolios()[TOP].getGeneCorrect():
Environment.getPortfolios()[TOP].getGeneCorrect()[gene] = average(
Environment.getPortfolios()[TOP].getGeneCorrect()[gene])
# plt.bar(range(len(Environment.getPortfolios()[TOP].getGeneCorrect())), Environment.getPortfolios()[TOP].getGeneCorrect().values(), align='center')
# plt.xticks(range(len(Environment.getPortfolios()[TOP].getGeneCorrect())), Environment.getPortfolios()[TOP].getGeneCorrect().keys())
# plt.show()
lsta = []
lstb = []
for gene in Environment.getPortfolios()[TOP].getGeneCorrect():
lsta.append(
Environment.getPortfolios()[TOP].getGeneCorrect()[gene])
lstb.append(Environment.getPortfolios()
[TOP].getSpecies().genes[gene].getWeight())
# plt.plot(lsta,lstb,"go")
print(np.corrcoef(lsta, lstb))
# plt.show()
#plt.plot(Environment.getPortfolios()[TOP].getShareCountList(),Environment.getPortfolios()[TOP].getOutputList(),"bo")
# return IndividualTester.Test(Environment.getPortfolios()[TOP].getSpecies().getGenes(), testStart, testEnd, shareCount, Investment),(Environment.getPortfolios()[TOP].getCorrectList().count(1)*1.0)/len(Environment.getPortfolios()[TOP].getCorrectList())
return IndividualTester.Test(Environment.getPortfolios()[TOP], testStart,
testEnd)
def Test(TopSpecies, testStart, testEnd):
for x in range(1):
try:
with open('bList.csv', 'rb') as csvfile:
balance = []
for row in csvfile:
balance.append(float(row))
Investment = balance[-1]
except(IndexError):
print "yikes"
Investment = 100000
stockList=["LUV"]
avg=[]
mx=[]
graph={}
g=[]
TOPV=0
Graph = []
Environment = Strategy(stockList)
numSpecies=1
pInvest=Investment
genCount=1
startDay= testStart
DAYS= testEnd
inputList={}
startPriceList={}
endPriceList={}
#Make Portfolios
Environment.addPortfolio("p0",pInvest)
Environment.getPortfolios()["p0"] = TopSpecies
Environment.getPortfolios()["p0"].reset(Investment)
#Set inputList
try:
inputList = pickle.load( open( "inputListA.p", "rb" ) )
startPriceList = pickle.load( open( "startPriceListA.p", "rb" ) )
endPriceList = pickle.load( open( "endPriceListA.p", "rb" ) )
except(IOError):
print "Writing New Pickle"
for d in range(startDay,DAYS):
if d in inputList:
print "Day", d, "Ready"
else:
print "Adding Day", d
Environment.setInputs("LUVBBF3", d, d+1)
inputList[d]=(Environment.getInputsF().copy())
startPriceList[d]=Environment.getRecentPrice()
endPriceList[d]=Environment.getEndPrice()
pickle.dump( inputList, open( "inputListA.p", "wb" ) )
pickle.dump( startPriceList, open( "startPriceListA.p", "wb" ) )
pickle.dump( endPriceList, open( "endPriceListA.p", "wb" ) )
#Make lists for randomization of initial population
inputs=inputList[startDay]
variables=inputs.keys()
right=[[],[]]
weight=[]
fitness={}
delimeters=["<",">"]
for valTop in np.arange(-1,1.1,.1):
right[0].append(round(valTop,2))
for y in variables:
right[1].append(y)
for z in np.arange(0,1,.01):
weight.append(round(z,3))
#Construct test population
# for portfolio in Environment.getPortfolios().keys():
# p = Environment.getPortfolios()[portfolio]
# p.addSpecies(inputs)
# for gene in TopSpecies:
# Environment.getPortfolios()['p0'].getSpecies().addGene(TopSpecies[gene].left,TopSpecies[gene].right, TopSpecies[gene].delimeter, TopSpecies[gene].weight)
# Start loop of days and generations
for day in range(startDay, DAYS):
inputs=inputList[day]
startPrice=startPriceList[day]
endPrice=endPriceList[day]
# Get boolean statements for the day
for portfolio in Environment.getPortfolios().keys():
for gene in Environment.getPortfolios()[portfolio].getSpecies().getGenes().keys():
Environment.getPortfolios()[portfolio].getSpecies().getGenes()[gene].makeBoolean(inputs)
# Get species output and add corresponding behavior
portfolio = "p0"
print Environment.getPortfolios()[portfolio].getSpecies().getRightThresh(),"right"
Environment.getPortfolios()[portfolio].addBehavior(Environment.getPortfolios()[portfolio].getSpecies().getOutput() > Environment.getPortfolios()[portfolio].getSpecies().getRightThresh() , ("Buy", "LUV",int(Environment.getPortfolios()[portfolio].getShareFactor()*(Environment.getPortfolios()[portfolio].getBalance()/(0.0+startPrice))),startPrice))
Environment.getPortfolios()[portfolio].addBehavior(Environment.getPortfolios()[portfolio].getSpecies().getOutput() < Environment.getPortfolios()[portfolio].getSpecies().getLeftThresh(), ("Short", "LUV",int(Environment.getPortfolios()[portfolio].getShareFactor()*(Environment.getPortfolios()[portfolio].getBalance()/(0.0+startPrice))),startPrice))
Environment.getPortfolios()[portfolio].addGeneCorrect(endPrice-startPrice)
#Go through with actions, sellBack at end of day and add balance to balanceList
Environment.getPortfolios()[portfolio].makeActions(startPrice)
Environment.getPortfolios()[portfolio].sellBack(stockList,endPrice)
Environment.getPortfolios()[portfolio].addBalance(Environment.getPortfolios()[portfolio].balance)
# for gene in Environment.getPortfolios()["p0"].getGeneCorrect():
# Environment.getPortfolios()["p0"].getGeneCorrect()[gene]=average(Environment.getPortfolios()["p0"].getGeneCorrect()[gene])
# plt.bar(range(len(Environment.getPortfolios()["p0"].getGeneCorrect())), Environment.getPortfolios()["p0"].getGeneCorrect().values(), align='center')
# plt.xticks(range(len(Environment.getPortfolios()["p0"].getGeneCorrect())), Environment.getPortfolios()["p0"].getGeneCorrect().keys())
# plt.show()
# lsta=[]
# lstb=[]
# for gene in Environment.getPortfolios()["p0"].getGeneCorrect():
# lsta.append(Environment.getPortfolios()["p0"].getGeneCorrect()[gene])
# lstb.append(Environment.getPortfolios()["p0"].getSpecies().genes[gene].getWeight())
# plt.plot(lsta,lstb,"go")
# print(np.corrcoef(lsta,lstb))
# plt.show()
#Add balanceList to graph
for portfolio in Environment.getPortfolios().keys():
fitness[portfolio]=Environment.getPortfolios()[portfolio].balanceList[-1]
sorted_fitness = sorted(fitness.items(), key=operator.itemgetter(1))
if(sorted_fitness[-1][-1]>TOPV):
TOP=sorted_fitness[-1][0]
TOPV=sorted_fitness[-1][1]
print "TEST DAY : ", DAYS
print "TEST CORRECT : ", (Environment.getPortfolios()["p0"].getCorrectList().count(1)*1.0)/len(Environment.getPortfolios()["p0"].getCorrectList())
print "TEST INCORRECT : ", (Environment.getPortfolios()["p0"].getCorrectList().count(-1)*1.0)/len(Environment.getPortfolios()["p0"].getCorrectList())
print "TEST HELD : ", (Environment.getPortfolios()["p0"].getCorrectList().count(0)*1.0)/len(Environment.getPortfolios()["p0"].getCorrectList())
print "TEST GUESSED : ", (Environment.getPortfolios()["p0"].getCorrectList().count(1)+Environment.getPortfolios()["p0"].getCorrectList().count(-1)*1.0)/len(Environment.getPortfolios()["p0"].getCorrectList())
print "LEFT THRESH : ", Environment.getPortfolios()[portfolio].getSpecies().getLeftThresh()
print "RIGHT THRESH : ", Environment.getPortfolios()[portfolio].getSpecies().getRightThresh()
print "SHARE FACTOR : ", Environment.getPortfolios()[portfolio].getShareFactor()
print "OUTPUT : ", Environment.getPortfolios()[portfolio].getSpecies().getOutput()
print "INVESTMENT : ", Investment
print "TEST BALANCE : ", Environment.getPortfolios()["p0"].getBalanceList()[-1]
print "DAY CHANGE : ", endPrice-startPrice
print "DAY PROFIT : ", Environment.getPortfolios()["p0"].getBalanceList()[-1]-Investment
return [(Environment.getPortfolios()["p0"].getCorrectList().count(1)*1.0)/len(Environment.getPortfolios()["p0"].getCorrectList()), Environment.getPortfolios()["p0"].getBalanceList()[-1]]
inputList = {}
startPriceList = {}
endPriceList = {}
#Make Portfolios
for individual in range(numSpecies):
Environment.addPortfolio("p" + str(individual), pInvest)
#Set inputList
for d in range(1, DAYS):
print "preparing for day", d
Environment.setInputs("LUVBBF3", d, d + 1)
inputList[d] = (Environment.getInputsF().copy())
startPriceList[d] = Environment.getRecentPrice()
endPriceList[d] = Environment.getEndPrice()
#Make lists for randomization of initial population
inputs = inputList[1]
variables = inputs.keys()
right = [[], []]
weight = []
fitness = {}
delimeters = ["<", ">"]
for valTop in np.arange(-1, 1.1, .1):
right[0].append(round(valTop, 2))
for y in variables:
right[1].append(y)
def Enviro(stockList, Investment, numSpecies, genCount, shareCount, trainStart,
trainEnd, testStart, testEnd, SaveNew, geneTest, evolutionPlot):
avg = []
mx = []
graph = {}
g = []
TOPV = 0
TOP = 0
Environment = Strategy(stockList)
pInvest = Investment
startDay = trainStart
DAYS = trainEnd
inputList = {}
startPriceList = {}
endPriceList = {}
#Make Portfolios
for individual in range(numSpecies):
Environment.addPortfolio("p" + str(individual), pInvest)
#Set inputList
try:
inputList = pickle.load(open("inputListA.p", "rb"))
startPriceList = pickle.load(open("startPriceListA.p", "rb"))
endPriceList = pickle.load(open("endPriceListA.p", "rb"))
except (IOError):
print "Writing New Pickle"
for d in range(startDay, DAYS):
if d in inputList:
print "Day", d, "Ready"
else:
print "Adding Day", d
Environment.setInputs("LUVBBF3", d, d + 1)
inputList[d] = (Environment.getInputsF().copy())
startPriceList[d] = Environment.getRecentPrice()
endPriceList[d] = Environment.getEndPrice()
pickle.dump(inputList, open("inputListA.p", "wb"))
pickle.dump(startPriceList, open("startPriceListA.p", "wb"))
pickle.dump(endPriceList, open("endPriceListA.p", "wb"))
#Make lists for randomization of initial population
inputs = inputList[startDay]
variables = inputs.keys()
right = [[], []]
weight = []
fitness = {}
delimeters = ["<", ">"]
for valTop in np.arange(-1, 1.1, .1):
right[0].append(round(valTop, 2))
for y in variables:
right[1].append(y)
for z in np.arange(0, 1, .01):
weight.append(round(z, 3))
#Construct initial population
for portfolio in Environment.getPortfolios().keys():
p = Environment.getPortfolios()[portfolio]
p.addSpecies(inputs)
for variable in Environment.getPortfolios()[portfolio].getSpecies(
).getGenes().keys():
a = random.randint(0, 1)
b = random.randint(0, len(right[a]) - 1)
c = delimeters[random.randint(0, 1)]
d = weight[random.randint(0, 99)]
Environment.getPortfolios()[portfolio].getSpecies().addGene(
variable, right[a][b], c, d)
Environment.getPortfolios()[portfolio].getSpecies().setThresh(
random.randrange(
-len(Environment.getPortfolios()
[portfolio].getSpecies().getGenes()), 1) + 0.0,
random.randrange(
0,
len(Environment.getPortfolios()
[portfolio].getSpecies().getGenes()) + 1) + 0.0)
Environment.getPortfolios()[portfolio].getSpecies().setShareFactor(
(random.randrange(1, 100)))
# Start loop of days and generations
for gen in range(genCount):
pool = mp.Pool(100)
partial_day_func = partial(MultiProcess1, Environment, inputList,
startPriceList, endPriceList, stockList)
pool.map(partial_day_func, [day for day in range(startDay, DAYS)])
pool.close()
pool.join()
#Get fitness of species
for portfolio in Environment.getPortfolios().keys():
# fitness[portfolio]=Environment.getPortfolios()[portfolio].balanceList[-1]
fitness[portfolio] = (Environment.getPortfolios(
)[portfolio].getCorrectList().count(1)) * 1.0 / len(
Environment.getPortfolios()[portfolio].getCorrectList())
#Find top 50%
sorted_fitness = sorted(fitness.items(), key=operator.itemgetter(1))
if (sorted_fitness[-1][-1] > TOPV):
TOP = sorted_fitness[-1][0]
TOPV = sorted_fitness[-1][1]
#Get bottom 50%
bottom = sorted_fitness[0:(numSpecies / 2)]
#Choose 50 pairs of top 50
top = sorted_fitness[numSpecies / 2:numSpecies]
topPair = []
for valTop in range(len(top)):
pair = valTop
while top[pair][0] == top[valTop][0]:
pair = random.randint(0, len(top) - 1)
topPair.append((top[valTop][0], top[pair][0]))
#For portfolio in 50 worst, set it equal to the breeding of a pair of top 50
for valBot in range(len(bottom)):
a = Environment.getPortfolios()[topPair[valBot]
[0]].getSpecies().copy()
b = Environment.getPortfolios()[topPair[valBot]
[1]].getSpecies().copy()
Environment.getPortfolios()[bottom[valBot][0]].resetAll(pInvest)
Environment.getPortfolios()[bottom[valBot][0]].updateSpecies(
a.breed(b).copy())
for valTop in range(len(top)):
Environment.getPortfolios()[top[valTop][0]].reset(pInvest)
print TOPV, "Fitness"
if (len(Environment.getPortfolios()[TOP].getCorrectList()) == 0):
print "bad things happend there was nothing there, uh hello"
else:
print "TRAIN CORRECT : ", (
Environment.getPortfolios()[TOP].getCorrectList().count(1) *
1.0) / len(Environment.getPortfolios()[TOP].getCorrectList())
print "TRAIN INCORRECT : ", (
Environment.getPortfolios()[TOP].getCorrectList().count(-1) *
1.0) / len(Environment.getPortfolios()[TOP].getCorrectList())
print "TRAIN HELD : ", (
Environment.getPortfolios()[TOP].getCorrectList().count(0) *
1.0) / len(Environment.getPortfolios()[TOP].getCorrectList())
print "TRAIN GUESSED : ", (
Environment.getPortfolios()[TOP].getCorrectList().count(1) +
Environment.getPortfolios()[TOP].getCorrectList().count(-1) *
1.0) / len(Environment.getPortfolios()[TOP].getCorrectList())
print "TRAIN BALANCE : ", Environment.getPortfolios(
)[TOP].getBalanceList()[-2]
# if evolutionPlot == True:
# plt.plot(Environment.getPortfolios()[TOP].getBalanceList())
# plt.show()
if geneTest == True:
for gene in Environment.getPortfolios()[TOP].getGeneCorrect():
Environment.getPortfolios()[TOP].getGeneCorrect()[gene] = average(
Environment.getPortfolios()[TOP].getGeneCorrect()[gene])
# plt.bar(range(len(Environment.getPortfolios()[TOP].getGeneCorrect())), Environment.getPortfolios()[TOP].getGeneCorrect().values(), align='center')
# plt.xticks(range(len(Environment.getPortfolios()[TOP].getGeneCorrect())), Environment.getPortfolios()[TOP].getGeneCorrect().keys())
# plt.show()
lsta = []
lstb = []
for gene in Environment.getPortfolios()[TOP].getGeneCorrect():
lsta.append(
Environment.getPortfolios()[TOP].getGeneCorrect()[gene])
lstb.append(Environment.getPortfolios()
[TOP].getSpecies().genes[gene].getWeight())
# plt.plot(lsta,lstb,"go")
print(np.corrcoef(lsta, lstb))
# plt.show()
#plt.plot(Environment.getPortfolios()[TOP].getShareCountList(),Environment.getPortfolios()[TOP].getOutputList(),"bo")
# return IndividualTester.Test(Environment.getPortfolios()[TOP].getSpecies().getGenes(), testStart, testEnd, shareCount, Investment),(Environment.getPortfolios()[TOP].getCorrectList().count(1)*1.0)/len(Environment.getPortfolios()[TOP].getCorrectList())
return IndividualTester.Test(
Environment.getPortfolios()[TOP].getSpecies().getGenes(), testStart,
testEnd, shareCount, Investment), (
Environment.getPortfolios()[TOP].getCorrectList().count(1) *
1.0) / len(Environment.getPortfolios()[TOP].getCorrectList())
def EnviroInitializer(stockList, Investment, numSpecies, genCount, shareCount,
startDay, endDay, testStart, testEnd, SaveNew, geneTest,
evolutionPlot, geneChange, oldStart, oldEnd):
avg = []
mx = []
graph = {}
g = []
TOPV = 0
TOP = 0
Environment = Strategy(stockList)
pInvest = Investment
startDay = startDay
DAYS = endDay
inputList = {}
startPriceList = {}
endPriceList = {}
inputListW = {}
startPriceListW = {}
endPriceListW = {}
#Make Portfolios
for individual in range(numSpecies):
Environment.addPortfolio("p" + str(individual), pInvest)
#Set Womb Battle days and womb change
try:
inputListW = pickle.load(open("inputListW1.p", "rb"))
startPriceListW = pickle.load(open("startPriceListW1.p", "rb"))
endPriceListW = pickle.load(open("endPriceListW1.p", "rb"))
except (IOError):
print "Writing New Pickle"
for d in range(oldStart, oldEnd):
if d in inputListW:
print "Old Day", d, "Ready"
else:
print "Adding Old Day", d
Environment.setInputs("LUVBBF3", d, d + 1)
inputListW[d] = (Environment.getInputsF().copy())
startPriceListW[d] = Environment.getRecentPrice()
endPriceListW[d] = Environment.getEndPrice()
pickle.dump(inputListW, open("inputListW1.p", "wb"))
pickle.dump(startPriceListW, open("startPriceListW1.p", "wb"))
pickle.dump(endPriceListW, open("endPriceListW1.p", "wb"))
#Set inputList
try:
inputList = pickle.load(open("inputListF1.p", "rb"))
startPriceList = pickle.load(open("startPriceListF1.p", "rb"))
endPriceList = pickle.load(open("endPriceListF1.p", "rb"))
except (IOError):
print "Writing New Pickle"
for d in range(startDay, DAYS):
if d in inputList:
print "Day", d, "Ready"
else:
print "Adding Day", d
Environment.setInputs("LUVBBF3", d, d + 1)
inputList[d] = (Environment.getInputsF().copy())
startPriceList[d] = Environment.getRecentPrice()
endPriceList[d] = Environment.getEndPrice()
pickle.dump(inputList, open("inputListF1.p", "wb"))
pickle.dump(startPriceList, open("startPriceListF1.p", "wb"))
pickle.dump(endPriceList, open("endPriceListF1.p", "wb"))
#Make lists for randomization of initial population
inputs = inputList[startDay]
variables = inputs.keys()
right = [[], []]
weight = []
fitness = {}
delimeters = ["<", ">"]
for valTop in np.arange(-1, 1.1, .1):
right[0].append(round(valTop, 2))
for y in variables:
right[1].append(y)
for z in np.arange(0, 1, .01):
weight.append(round(z, 3))
#Construct initial population
for portfolio in Environment.getPortfolios().keys():
p = Environment.getPortfolios()[portfolio]
p.addSpecies(inputs)
for variable in Environment.getPortfolios()[portfolio].getSpecies(
).getGenes().keys():
a = random.randint(0, 1)
b = random.randint(0, len(right[a]) - 1)
c = delimeters[random.randint(0, 1)]
d = weight[random.randint(0, 99)]
Environment.getPortfolios()[portfolio].getSpecies().addGene(
variable, right[a][b], c, d)
Environment.getPortfolios()[portfolio].getSpecies().setThresh(
random.randrange(
-len(Environment.getPortfolios()
[portfolio].getSpecies().getGenes()), 1) + 0.0,
random.randrange(
0,
len(Environment.getPortfolios()
[portfolio].getSpecies().getGenes()) + 1) + 0.0)
Environment.getPortfolios()[portfolio].getSpecies().setShareFactor(
(random.randrange(1, 100)))
return inputList, startPriceList, endPriceList, Environment
def Enviro(stockList,Investment,numSpecies, genCount, shareCount, startDay, endDay, testStart, testEnd, SaveNew, geneTest, evolutionPlot, geneChange, oldStart,oldEnd):
avg=[]
mx=[]
graph={}
g=[]
TOPV=0
TOP=0
Environment = Strategy(stockList)
pInvest=Investment
startDay = startDay
DAYS=endDay
inputList={}
startPriceList={}
endPriceList={}
inputListW = {}
startPriceListW = {}
endPriceListW = {}
#Make Portfolios
for individual in range(numSpecies):
Environment.addPortfolio("p"+str(individual),pInvest)
#Set Womb Battle days and womb change
try:
inputListW = pickle.load( open( "inputListW1.p", "rb" ) )
startPriceListW = pickle.load( open( "startPriceListW1.p", "rb" ) )
endPriceListW = pickle.load( open( "endPriceListW1.p", "rb" ) )
except(IOError):
print "Writing New Pickle"
for d in range(oldStart,oldEnd):
if d in inputListW:
print "Old Day", d, "Ready"
else:
print "Adding Old Day", d
Environment.setInputs("LUVBBF3", d, d+1)
inputListW[d]=(Environment.getInputsF().copy())
startPriceListW[d]=Environment.getRecentPrice()
endPriceListW[d]=Environment.getEndPrice()
pickle.dump( inputListW, open( "inputListW1.p", "wb" ) )
pickle.dump( startPriceListW, open( "startPriceListW1.p", "wb" ) )
pickle.dump( endPriceListW, open( "endPriceListW1.p", "wb" ) )
#Set inputList
try:
inputList = pickle.load( open( "inputListF1.p", "rb" ) )
startPriceList = pickle.load( open( "startPriceListF1.p", "rb" ) )
endPriceList = pickle.load( open( "endPriceListF1.p", "rb" ) )
except(IOError):
print "Writing New Pickle"
for d in range(startDay,DAYS):
if d in inputList:
print "Day", d, "Ready"
else:
print "Adding Day", d
Environment.setInputs("LUVBBF3", d, d+1)
inputList[d]=(Environment.getInputsF().copy())
startPriceList[d]=Environment.getRecentPrice()
endPriceList[d]=Environment.getEndPrice()
pickle.dump( inputList, open( "inputListF1.p", "wb" ) )
pickle.dump( startPriceList, open( "startPriceListF1.p", "wb" ) )
pickle.dump( endPriceList, open( "endPriceListF1.p", "wb" ) )
#Make lists for randomization of initial population
inputs=inputList[startDay]
variables=inputs.keys()
right=[[],[]]
weight=[]
fitness={}
delimeters=["<",">"]
for valTop in np.arange(-1,1.1,.1):
right[0].append(round(valTop,2))
for y in variables:
right[1].append(y)
for z in np.arange(0,1,.01):
weight.append(round(z,3))
#Construct initial population
for portfolio in Environment.getPortfolios().keys():
p = Environment.getPortfolios()[portfolio]
p.addSpecies(inputs)
for variable in Environment.getPortfolios()[portfolio].getSpecies().getGenes().keys():
a=random.randint(0,1)
b=random.randint(0,len(right[a])-1)
c=delimeters[random.randint(0,1)]
d=weight[random.randint(0,99)]
Environment.getPortfolios()[portfolio].getSpecies().addGene(variable,right[a][b],c,d)
Environment.getPortfolios()[portfolio].getSpecies().setThresh(random.randrange(-len(Environment.getPortfolios()[portfolio].getSpecies().getGenes()),1)+0.0,random.randrange(0,len(Environment.getPortfolios()[portfolio].getSpecies().getGenes())+1)+0.0)
Environment.getPortfolios()[portfolio].getSpecies().setShareFactor((random.randrange(1,100)))
# Start loop of days and generations
Corro = []
for gen in range(genCount):
for day in range(startDay,DAYS):
print "day " + str(day) +" year "+ str(gen+1)
inputs=inputList[day]
startPrice=startPriceList[day]
endPrice=endPriceList[day]
# Get boolean statements for the day
for portfolio in Environment.getPortfolios().keys():
for gene in Environment.getPortfolios()[portfolio].getSpecies().getGenes().keys():
Environment.getPortfolios()[portfolio].getSpecies().getGenes()[gene].makeBoolean(inputs)
# Get species output and add corresponding behavior
for portfolio in Environment.getPortfolios().keys():
Environment.getPortfolios()[portfolio].addBehavior(Environment.getPortfolios()[portfolio].getSpecies().getOutput() > Environment.getPortfolios()[portfolio].getSpecies().getRightThresh() , ("Buy", "LUV",int(Environment.getPortfolios()[portfolio].getShareFactor()*(Environment.getPortfolios()[portfolio].getBalance()/(0.0+startPrice))),startPrice))
Environment.getPortfolios()[portfolio].addBehavior(Environment.getPortfolios()[portfolio].getSpecies().getOutput() < Environment.getPortfolios()[portfolio].getSpecies().getLeftThresh(), ("Short", "LUV",int(Environment.getPortfolios()[portfolio].getShareFactor()*(Environment.getPortfolios()[portfolio].getBalance()/(0.0+startPrice))),startPrice))
Environment.getPortfolios()[portfolio].addGeneCorrect(endPrice-startPrice)
#Go through with actions, sellBack at end of day and add balance to balanceList
for portfolio in Environment.getPortfolios().keys():
Environment.getPortfolios()[portfolio].makeActions(startPrice)
Environment.getPortfolios()[portfolio].sellBack(stockList,endPrice)
Environment.getPortfolios()[portfolio].addBalance(Environment.getPortfolios()[portfolio].balance)
#Get fitness of species
for portfolio in Environment.getPortfolios().keys():
# fitness[portfolio]=Environment.getPortfolios()[portfolio].balanceList[-1]
fitness[portfolio]=(Environment.getPortfolios()[portfolio].getCorrectList().count(1))*1.0/len(Environment.getPortfolios()[portfolio].getCorrectList())
#Find top 50%
sorted_fitness = sorted(fitness.items(), key=operator.itemgetter(1))
if(sorted_fitness[-1][-1]>TOPV):
TOP=sorted_fitness[-1][0]
TOPV=sorted_fitness[-1][1]
#Get bottom 50%
bottom=sorted_fitness[0:(numSpecies/2)]
#Choose 50 pairs of top 50
top=sorted_fitness[numSpecies/2:numSpecies]
topPair=[]
for valTop in range(len(top)):
pair=valTop
while top[pair][0]==top[valTop][0]:
pair=random.randint(0,len(top)-1)
topPair.append((top[valTop][0],top[pair][0]))
#For portfolio in 50 worst, set it equal to the breeding of a pair of top 50
for valBot in range(len(bottom)):
a = Environment.getPortfolios()[topPair[valBot][0]].getSpecies().copy()
b=Environment.getPortfolios()[topPair[valBot][1]].getSpecies().copy()
Environment.getPortfolios()[bottom[valBot][0]].resetAll(pInvest)
Environment.getPortfolios()[bottom[valBot][0]].updateSpecies(a.breed(b).copy())
for valTop in range(len(top)):
Environment.getPortfolios()[top[valTop][0]].reset(pInvest)
plt.plot(Environment.getPortfolios()[TOP].getBalanceList())
if geneChange == True:
lsta=[]
lstb=[]
for gene in Environment.getPortfolios()[TOP].getGeneCorrect():
lsta.append(average(Environment.getPortfolios()[TOP].getGeneCorrect()[gene]))
lstb.append(Environment.getPortfolios()[TOP].getSpecies().genes[gene].getWeight())
print(np.corrcoef(lsta,lstb)[0][1])
Corro.append(np.corrcoef(lsta,lstb)[0][1])
print TOPV, "Fitness"
print "SPECIES : ", Environment.getPortfolios()[TOP].getSpecies().printSpecies()
if len(Environment.getPortfolios()[TOP].getCorrectList()) == 0:
print "IT HAPPENED"
else:
print "CORRECT : ", (Environment.getPortfolios()[TOP].getCorrectList().count(1)*1.0)/len(Environment.getPortfolios()[TOP].getCorrectList())
print "INCORRECT : ", (Environment.getPortfolios()[TOP].getCorrectList().count(-1)*1.0)/len(Environment.getPortfolios()[TOP].getCorrectList())
print "HELD : " , (Environment.getPortfolios()[TOP].getCorrectList().count(0)*1.0)/len(Environment.getPortfolios()[TOP].getCorrectList())
print "GUESSED : " , (Environment.getPortfolios()[TOP].getCorrectList().count(1)+Environment.getPortfolios()[TOP].getCorrectList().count(-1)*1.0)/len(Environment.getPortfolios()[TOP].getCorrectList())
print "BALANCE : ", Environment.getPortfolios()[TOP].getBalanceList()[-2]
plt.show()
if geneChange == True:
plt.plot(range(genCount), Corro, 'go')
plt.show()
if evolutionPlot == True:
plt.plot(Environment.getPortfolios()[TOP].getBalanceList())
plt.show()
if geneTest == True:
for gene in Environment.getPortfolios()[TOP].getGeneCorrect():
Environment.getPortfolios()[TOP].getGeneCorrect()[gene]=average(Environment.getPortfolios()[TOP].getGeneCorrect()[gene])
plt.bar(range(len(Environment.getPortfolios()[TOP].getGeneCorrect())), Environment.getPortfolios()[TOP].getGeneCorrect().values(), align='center')
plt.xticks(range(len(Environment.getPortfolios()[TOP].getGeneCorrect())), Environment.getPortfolios()[TOP].getGeneCorrect().keys())
locs, labels = plt.xticks()
plt.setp(labels, rotation=90)
plt.show()
lsta=[]
lstb=[]
for gene in Environment.getPortfolios()[TOP].getGeneCorrect():
lsta.append(Environment.getPortfolios()[TOP].getGeneCorrect()[gene])
lstb.append(Environment.getPortfolios()[TOP].getSpecies().genes[gene].getWeight())
print "CORRELATION :", np.corrcoef(lsta,lstb)[0][1]
plt.plot(lsta,lstb,"go")
plt.show()
print "Testing Robustness..."
IndividualTester.Test(Environment.getPortfolios()[TOP].getSpecies().getGenes(), testStart, testEnd, shareCount, Investment, SaveNewTest)
