def __init__(self, genome, initValue):
# print("NEW MODEL")
self.__genome = genome
self.__brain = []
self.__layerMap = []
# initValue refers to the amount of money we are willing to commit to a single trading strategy
self.__initValue = initValue
self.__equity = initValue
self.__stocks = [0, 0, 0, 0, 0]
self.__lastPrice = [0.0, 0.0, 0.0, 0.0, 0.0]
# note that model networks now have their neurons divided into layers and that genes are translated into neurons
self.__inputLayer = []
self.__outputLayer = []
for g in self.__genome.genes():
if g.isEnabled() == True:
newNeuronA = neatneuron.Neuron(g.inputNum())
newNeuronB = neatneuron.Neuron(g.outputNum())
if newNeuronA not in self.__brain:
newNeuronA.addNout(neatneuron.Connection(g.outputNum(), g.weight()))
self.__brain.append(newNeuronA)
if newNeuronA.node() in self.__genome.inputSet():
self.__inputLayer.append(newNeuronA)
if newNeuronA.node() in self.__genome.outputSet():
self.__outputLayer.append(newNeuronA)
else:
self.__brain[self.__brain.index(newNeuronA)].addNout(neatneuron.Connection(g.outputNum(), g.weight()))
if newNeuronB not in self.__brain:
self.__brain.append(newNeuronB)
if newNeuronB.node() in self.__genome.outputSet():
self.__outputLayer.append(newNeuronB)
if newNeuronB.node() in self.__genome.inputSet():
self.__inputLayer.append(newNeuronB)
self.__layerMap.append(self.__inputLayer)
self.separateIntoLayers()
self.__layerMap.append(self.__outputLayer)
def runTraced(self, data, stock, point, file):
for d in data:
position = point * 5 + data.index(d)
# Data Point d - set of attributes that are deemed necessary for trading strategies
for neuron in self.__brain:
neuron.resetVal()
self.__lastPrice[stock] = d[0]
for i in range(0, len(d)):
if i < len(self.__inputLayer):
self.__inputLayer[i].resetVal()
self.__inputLayer[i].passVal(d[i])
for layer in self.__layerMap:
for neuron in layer:
for output in neuron.nouts():
nextNeuron = self.__brain[self.__brain.index(neatneuron.Neuron(output.output))]
nextNeuron.passVal(activation_function(neuron.val() * output.weight))
for neuronIndex in range(len(self.__outputLayer)):
if self.__outputLayer[neuronIndex].val() > 1.0:
buyQuant = self.__stocks[stock]
self.buy(stock, self.__outputLayer[neuronIndex].val() - 1.0)
buyQuant -= self.__stocks[stock]
buyQuant *= -1
if buyQuant > 0:
file.write(str(position) + "," + str(buyQuant) + "," + str(stock) + "\n")
if buyQuant < 0:
file.write(str(position) + "," + str(buyQuant) + "," + str(stock) + "\n")
if self.__outputLayer[neuronIndex].val() < -1.0:
#print("SELL FUNCTION CALLED")
sellQuant = self.__stocks[stock]
self.sell(stock, -1 * (self.__outputLayer[neuronIndex].val() + 1.0))
sellQuant -= self.__stocks[stock]
file.write(str(position) + "," + str(sellQuant) + "," + str(stock) + "\n")
def separateIntoLayers(self):
done = False
while done == False:
tentativeLayer = []
for no in self.__layerMap[-1]:
for out in no.nouts():
outn = self.__brain[self.__brain.index(neatneuron.Neuron(out.output))]
if outn not in self.__outputLayer and outn not in tentativeLayer:
if outn.layer() == -1 or outn.layer() > no.layer():
tentativeLayer.append(outn)
outn.setLayer(len(self.__layerMap))
if len(tentativeLayer) > 0:
self.__layerMap.append(tentativeLayer)
else:
done = True
def runAlt(self, data, stock):
for d in data:
# Data Point d - set of attributes that are deemed necessary for trading strategies
for neuron in self.__brain:
neuron.resetVal()
self.__lastPrice[stock] = d[0]
for i in range(0, len(d)):
if i < len(self.__inputLayer):
self.__inputLayer[i].resetVal()
self.__inputLayer[i].passVal(d[i])
for layer in self.__layerMap:
for neuron in layer:
for output in neuron.nouts():
nextNeuron = self.__brain[self.__brain.index(neatneuron.Neuron(output.output))]
nextNeuron.passVal(activation_function(neuron.val() * output.weight))
for neuronIndex in range(len(self.__outputLayer)):
if self.__outputLayer[neuronIndex].val() > 1.0:
self.buy(stock, self.__outputLayer[neuronIndex].val() - 1.0)
if self.__outputLayer[neuronIndex].val() < -1.0:
#print("SELL FUNCTION CALLED")
self.sell(stock, -1 * (self.__outputLayer[neuronIndex].val() + 1.0))