def predictNextValue(self):
if self.last_day is None:
print "No data in model!"
return
stockData, x_data, y_data = self.__getStockData(to_date = csvmgr.stockDayToDatetime(self.last_day))
print "Prediction: Value for "+self.csv_id+" will "\
+("increase" if self.model.predict_classes(x_data, batch_size=self.batch_size)[-1][0] else "decrease")
def feedWithStockData(self, limit_date = None, fresh = False):
if self.last_day is None:
stockData, x_data, y_data = self.__getStockData(to_date = limit_date, fresh=fresh)
else:
last_date = csvmgr.stockDayToDatetime(self.last_day)
last_date += timedelta(1)
stockData, x_data, y_data = self.__getStockData(from_date = last_date, to_date = limit_date, fresh=fresh)
print "Fitting..."
self.last_day = csvmgr.getLastDayFromStock(stockData)
def feedWithStockData(self, limit_date = None, fresh = False):
if self.last_day is None:
stockData, x_data, y_data = self.__getStockData(to_date = limit_date, fresh=fresh)
else:
last_date = csvmgr.stockDayToDatetime(self.last_day)
last_date += timedelta(1)
stockData, x_data, y_data = self.__getStockData(from_date = last_date, to_date = limit_date, fresh=fresh)
print "Fitting..."
self.model.fit(x_data, y_data, batch_size=self.batch_size, nb_epoch=self.nb_epoch,
show_accuracy=True)
self.last_day = csvmgr.getLastDayFromStock(stockData)
def __getStockData(self, from_date=None, to_date=None, fresh=False):
maxlen = self.days_seq
stockData = csvmgr.getStockData(self.csv_id, force_update=fresh)
if from_date is None:
from_date = csvmgr.stockDayToDatetime(csvmgr.getFirstDayFromStock(stockData))
if to_date is None:
to_date = csvmgr.stockDayToDatetime(csvmgr.getLastDayFromStock(stockData))
if (to_date-from_date) <= timedelta(maxlen):
from_date -= timedelta(maxlen+20)
stockData = csvmgr.filterStockData(stockData, from_date, to_date)
print "Getting stock data from "+csvmgr.getFirstDayFromStock(stockData) + " to " + csvmgr.getLastDayFromStock(stockData)
stockValues = csvmgr.getPricesFromStock(stockData)
x_data = numpy.array([stockValues[i:i+maxlen] for i in xrange(len(stockValues)-maxlen)])
y_data = numpy.array([floor(stockValues[i]/stockValues[i-1]) for i in xrange(maxlen,len(stockValues))])
return stockData, x_data, y_data
parser_predict = subparsers.add_parser("predict", help="Predict growth of stock prices")
parser_predict.add_argument("name", nargs=1, help="Name of used prediction model")
print("")
args = parser.parse_args()
if args.command == "create":
model = predictors[args.type[0]](args.name[0])
if args.description == "":
args.description = [args.name[0] + " model"]
model.createModel(args.stock_id[0], args.description[0])
if args.limitdate != "":
model.feedWithStockData(limit_date=stockDayToDatetime(args.limitdate))
else:
model.feedWithStockData(fresh=True)
model.saveModel()
elif args.command == "update":
model = modelLoader(args.name[0])
if model is None:
print "Model "+args.name[0]+" doesn't exist!"
else:
model.feedWithStockData(fresh=True)
model.saveModel()
elif args.command == "print":
model = modelLoader(args.name[0], load_model=False)
if model is None:
print "Model "+args.name[0]+" doesn't exist!"
elif args.command == "test":