def ANNCaller(stockSymbol, isDaily=True):
dataTrain = readJson(stockSymbol, 10, isDaily)
# print(dataTrain)
myNN = NN(11, 5, 1)
myNN.loadWeights(stockSymbol)
result = []
resultJson = {}
resultJson[stockSymbol] = {}
prediction = [0] * 4
for i in range(4):
# print(dataTrain[i])
prediction[i] = round(myNN.runNN(dataTrain[i][::-1])[0], 2)
# print(dataTrain)
stdev = round(statistics.stdev(dataTrain[3]), 2)
averageErrorPath = ''
if isDaily:
averageErrorPath = path.abspath(
path.join(
path.abspath(sys.argv[0]),
"..")) + '\\dataset\\daily\\prediction_averageErr_ANN.json'
else:
averageErrorPath = path.abspath(
path.join(
path.abspath(sys.argv[0]),
"..")) + '\\dataset\\intraday\\prediction_averageErr_ANNA.json'
with open(averageErrorPath, 'r') as f:
averageError = json.load(f)[stockSymbol]
# print(prediction)
return (prediction[0], prediction[1], prediction[2], prediction[3],
averageError, stdev)
def bayesianCaller(stockSymbol, isDaily=True):
n = 20
m = 6
date = 0
# fileArray=['NFLX.csv', 'NVDA.csv', 'TESL.csv', 'WB.csv']
# dataSource=readCSV(stockSymbol,dataType,date,n)
dataTrain = readJson(stockSymbol, n, isDaily)
# readCSV(stockSymbol,dataType,date,n)
# if date<0 or date+n>=len(dataSource):
# return [-2,-2,-2]
# print(dataTrain)
if date + n > len(dataTrain[0]):
return ([-2, -2, -2, -2], -2, -2)
# dataTrain=dataSource[date:date+n][::-1]
# for n in [6,10,16,20]:
# print('n='+str(n))
# for m in [5,6,7,8]:
# print('n='+str(n)+', m='+str(m))
# result=[]
# avgRelErr=0
# for stockSymbol in fileArray:
# for pointStart in [1,101,201]:
prediction = [0] * 4
for i in range(4):
# print(dataTrain[i])
prediction[i] = round(bayesian(dataTrain[i], m, n), 2)
# print(dataTrain)
stdev = round(statistics.stdev(dataTrain[3]), 2)
averageErrorPath = ''
if isDaily:
#ctypes.windll.user32.MessageBoxW(0, path.abspath(path.join(path.abspath(sys.argv[0]), "..")), "Your title", 1)
averageErrorPath = path.abspath(
path.join(path.abspath(sys.argv[0]), "..")
) + '\\dataset\\daily\\prediction_averageErr_bayesian.json'
else:
#ctypes.windll.user32.MessageBoxW(0, path.abspath(path.join(path.abspath(sys.argv[0]), "..")), "Your title", 1)
averageErrorPath = path.abspath(
path.join(path.abspath(sys.argv[0]), "..")
) + '\\dataset\\intraday\\prediction_averageErr_bayesian.json'
with open(averageErrorPath, 'r') as f:
averageError = json.load(f)[stockSymbol]
# print(prediction)
return (prediction[0], prediction[1], prediction[2], prediction[3],
averageError, stdev)
def callerANN(stockSymbol, isDaily=True):
dataSource = readJson(stockSymbol, 61, isDaily)
# print(dataSource)
global stock
stock = stockSymbol
result = []
# prediction=[0]*4
# print(dataSource[3][10:])
dataTrain = dataFormatter(dataSource[3][10:][::-1], 10, 5, 51)[::-1]
dateArray = []
for num in [10, 20, 30, 40, 50]:
dateArray.append(dataSource[4][num])
# print(dateArray)
# print(dataTrain)
myNN = NN(11, 5, 1)
# myNN.initWeightsRand()
myNN.loadWeights()
for inputData, outputData in dataTrain:
predictionItem = myNN.runNN(inputData)
print(inputData, outputData, predictionItem)
def bayesianHistory(stockSymbol, dateScope, gap, isDaily=True):
m = 6
n = gap
dataSource = readJson(stockSymbol, -1, isDaily)[3:5] #price close only
# print(dataSource)
# readCSV(stockSymbol,dataType,date,n)
# if date<0 or date+n>=len(dataSource):
# return [-2,-2,-2]
if dateScope >= len(dataSource[0]):
return -1
errorSum = 0
result = {}
prediction = bayesian(dataSource[0][9::-1], m, n)
result['nextDay'] = {}
result['nextDay']['prediction'] = prediction
for k in range(gap, dateScope - gap, gap):
dataTrain = dataSource[0][k:k + gap][::-1]
trueValue = dataSource[0][k - 1]
date = dataSource[1][k - 1]
prediction = bayesian(dataTrain, m, n)
error = abs((prediction - trueValue) / trueValue * 100)
result[date] = {}
result[date]['prediction'] = prediction
result[date]['trueValue'] = trueValue
result[date]['relativeError'] = error
errorSum += error
# for n in [6,10,16,20]:
# print('n='+str(n))
# for m in [5,6,7,8]:
# print('n='+str(n)+', m='+str(m))
# result=[]
# avgRelErr=0
# for stockSymbol in fileArray:
# for pointStart in [1,101,201]:
# prediction=[0]*4
# for i in range(4):
# prediction[i] = round(bayesian(dataTrain[i],m,n),2)
# # print(dataTrain)
# stdev=round(statistics.stdev(dataTrain[3]),2)
# # print(prediction)
# return(prediction,-1,stdev)
# print (result)
averageError = errorSum / len(result)
# print(averageError)
curvefile = ''
avgFile = ''
if isDaily:
curvefile = 'dataset/daily/prediction_' + stockSymbol + '_bayesian.json'
avgFile = 'dataset/daily/prediction_averageErr_bayesian.json'
else:
curvefile = 'dataset/intraday/prediction_' + stockSymbol + '_bayesian.json'
avgFile = 'dataset/intraday/prediction_averageErr_bayesian.json'
# print(result)
with open(curvefile, "w") as f:
json.dump(result, f)
with open(avgFile, "r") as f:
# errStr=json.load(f)
errDict = json.load(f)
errDict[stockSymbol] = averageError
with open(avgFile, "w") as f:
json.dump(errDict, f)
def testANN(stockSymbol, isDaily=True):
dataSource = readJson(stockSymbol, 250, isDaily)
# print(dataSource)
global stock
stock = stockSymbol
myNN = NN(11, 5, 1)
# myNN.initWeightsRand()
myNN.loadWeights(stockSymbol)
result = []
resultJson = {}
resultJson[stockSymbol] = {}
# prediction=[0]*4
# print(dataSource[3][10:])
dataTrain = dataFormatter(dataSource[3][9:][::-1], 10, 5, 241)[::-1]
# print(len(dataTrain))
dateArray = []
for num in range(10, 241, 10):
dateArray.append(dataSource[4][num])
errorSum = 0
result = {}
# print(dataSource[3][:10][::-1])
prediction = myNN.runNN(dataSource[3][:10][::-1])[0]
# print(prediction)
result['nextDay'] = {}
result['nextDay']['prediction'] = prediction
for i in range(24):
dataItem = dataTrain[i]
# print(dataTrain)
date = dateArray[i]
# dataTrain=dataTrain[0]
trueValue = dataItem[1][0]
# print(dataTrain[0])
prediction = myNN.runNN(dataItem[0])[0]
error = abs((prediction - trueValue) / trueValue * 100)
result[date] = {}
result[date]['prediction'] = prediction
result[date]['trueValue'] = trueValue
result[date]['relativeError'] = error
errorSum += error
# for n in [6,10,16,20]:
# print('n='+str(n))
# for m in [5,6,7,8]:
# print('n='+str(n)+', m='+str(m))
# result=[]
# avgRelErr=0
# for stockSymbol in fileArray:
# for pointStart in [1,101,201]:
# prediction=[0]*4
# for i in range(4):
# prediction[i] = round(bayesian(dataTrain[i],m,n),2)
# # print(dataTrain)
# stdev=round(statistics.stdev(dataTrain[3]),2)
# # print(prediction)
# return(prediction,-1,stdev)
# print (result)
averageError = errorSum / len(result)
# print(averageError)
curvefile = ''
avgFile = ''
if isDaily:
curvefile = path.abspath(
path.join(path.abspath(sys.argv[0]), "..")
) + '\\dataset\\daily\\prediction_' + stockSymbol + '_ANN.json'
avgFile = path.abspath(path.join(
path.abspath(sys.argv[0]),
"..")) + '\\dataset\\daily\\prediction_averageErr_ANN.json'
else:
curvefile = path.abspath(
path.join(path.abspath(sys.argv[0]), "..")
) + '\\dataset\\intraday\\prediction_' + stockSymbol + '_ANN.json'
avgFile = path.abspath(path.join(
path.abspath(sys.argv[0]),
"..")) + '\\dataset\\intraday\\prediction_averageErr_ANN.json'
# print(result)
with open(curvefile, "w") as f:
json.dump(result, f)
with open(avgFile, "r") as f:
# errStr=json.load(f)
errDict = json.load(f)
errDict[stockSymbol] = averageError
with open(avgFile, "w") as f:
json.dump(errDict, f)
# print(dateArray)
# print(dataTrain)
for inputData, outputData in dataTrain:
predictionItem = myNN.runNN(inputData)
print(inputData, outputData, predictionItem)