import Formatter import ANN as nn import numpy as np train_period = 24 * 7 # 7 days test_period = 24 # 1 day bin_count = 8 period_class = Formatter.PeriodSample(60) target = [] change_data = [] matrix = [ period_class.getChangeVolData(train_period, test_period) for i in range(900000) ] for index in range(0, len(matrix), 1): change_data.append(matrix[index][0][:, 1]) bin_no = np.zeros([bin_count], dtype=float) bin_no[matrix[index][1]] = 1.0 target.append(bin_no) change_data = np.array(change_data, dtype=float) model = nn.ann() cost = model.train(change_data, target) # plt.plot(cost) print(cost) # plt.show() # print(cost) print((model.test(change_data, target)))
import numpy as np from keras.callbacks import TensorBoard, ModelCheckpoint from keras.layers import Dense, LSTM, Merge from keras.models import Sequential, model_from_json from keras.optimizers import RMSprop import keras bin_count = 8 # no of bins BATCH_SIZE = 1000 import Formatter period_sample = Formatter.PeriodSample(60) INPUT_SIZE = 24 def createModel(train_period, target): cost = RMSprop(lr=0.001, rho=0.9, epsilon=None, decay=0.0) EMA_lstm = Sequential() EMA_lstm.add( LSTM(INPUT_SIZE, input_shape=(INPUT_SIZE, 1), batch_input_shape=(BATCH_SIZE, INPUT_SIZE, 1), dropout=0.2, return_sequences=False)) K_lstm = Sequential() K_lstm.add( LSTM(INPUT_SIZE, input_shape=(INPUT_SIZE, 1), batch_input_shape=(BATCH_SIZE, INPUT_SIZE, 1),
import Data import Formatter import grouping_changecalc as grp print(Data.getNames()) formatter = Formatter.PeriodSample(10) formatter.getChangeVolData(50, 5) # # # train, test = Data.randomSample(10,2) # print(Data.getNames())