def __init__(self,
file_name=None,
l_fac=1.0,
s_fac=1.0,
year_range=range(2006, 2021),
l_range=range(12, 31, 2),
s_range=range(2, 12)):
# 在此修改长短期均线长度
# ========================
self.l_range = l_range # 长期均线选择
self.s_range = s_range # 短期均线选择
self.year_range = year_range
# ============================
# 需要输入文件名
assert file_name != None
# 载入文件
self.close_data = load_file(file_name).read_file()
# year range从loadfile中传递？
self.time_list = self.get_ftd_list()
# print(self.time_list)
self.ret_list = {}
# MACD指标选择的长短期为12,26
self.l_fac = l_fac
self.s_fac = s_fac
self.main_loop()
def __init__(self,
file_name=None,
l_fac=1.0,
s_fac=1.06,
year_range=range(2006, 2021),
lma=14,
sma=3):
# 在此修改长短期均线长度
# ========================
self.lma = lma # 长期均线选择
self.sma = sma # 短期均线选择
self.year_range = year_range
# ============================
# 需要输入文件名
assert file_name != None
# 载入文件
self.close_data = load_file(file_name).read_file()
# year range从loadfile中传递？
# print(self.time_list)
self.ret_list = {}
# MACD指标选择的长短期为12,26
self.l_fac = l_fac
self.s_fac = s_fac
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error
from LoadFile import load_file
import os
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
DATA_FILE='./data/CSI.csv'
MODEL_NAME='my_model'
# 读取数据
close_earning=load_file(DATA_FILE).read_file()
dataset=close_earning['earning'].values.reshape(-1,1)
# data = read_csv('r.csv') # csv文件 n*1 ,n代表样本数，反应时间序列，1维数据
#
# values1 = data.values;
# dataset = values1[:, 0].reshape(-1, 1) # 注意将一维数组，转化为2维数组
# dataset = dataset.astype('float32') # 将数据转化为32位浮点型，防止0数据
def cal_the_return(testPredict,testY):
compare=pd.DataFrame(testPredict,)
compare['testy']=testY[0]
# 测试集的实际收益率和测试集的预测收益率
compare.columns=['predy','testy']
compare.to_csv('compare.csv')
accy = [1]
accp = [[1]for i in range(6)]
def initial_model(self):
# 读取数据
close_earning = load_file(self.data_file).read_file()
dataset = close_earning['earning'].values.reshape(-1, 1)
# data = read_csv('r.csv') # csv文件 n*1 ,n代表样本数，反应时间序列，1维数据
#
# values1 = data.values;
# dataset = values1[:, 0].reshape(-1, 1) # 注意将一维数组，转化为2维数组
# dataset = dataset.astype('float32') # 将数据转化为32位浮点型，防止0数据
numpy.random.seed(7) # 随机数生成时算法所用开始的整数值
# normalize the dataset
scaler = MinMaxScaler(feature_range=(0, 1)) # 归一化0-1
dataset = scaler.fit_transform(dataset)
# split into train and test sets #训练集和测试集分割
train_size = int(len(dataset) * 0.67) # %67的训练集，剩下测试集
test_size = len(dataset) - train_size
train, test = dataset[0:train_size, :], dataset[train_size:len(
dataset), :] # 训练集和测试集
# use this function to prepare the train and test datasets for modeling
trainX, trainY = self.create_dataset(train) # 训练输入输出
testX, testY = self.create_dataset(test) # 测试输入输出
# reshape input to be [samples, time steps, features]#注意转化数据维数
trainX = numpy.reshape(trainX, (trainX.shape[0], 1, trainX.shape[1]))
testX = numpy.reshape(testX, (testX.shape[0], 1, testX.shape[1]))
# def mean_squared_error(y_true, y_pred):
# return K.mean(K.square(y_pred - y_true), axis=-1)
if self.train_flag:
# 建立LSTM模型
model = Sequential()
model.add(LSTM(
11,
input_shape=(1, self.look_back))) # 隐层11个神经元 （可以断调整此参数提高预测精度）
model.add(Dense(1))
model.compile(loss=self.loss,
optimizer=self.optimizer) # 评价函数mse，优化器adam
model.fit(trainX,
trainY,
epochs=self.epoch,
batch_size=100,
verbose=2) # 100次迭代
# save the model
# model.save_weights("my_model_weights.h5") # only save the weight
model.save('./model_file/{}.h5'.format(self.model_name))
else:
model = load_model('./model_file/{}.h5'.format(self.model_name))
trainPredict = model.predict(trainX)
testPredict = model.predict(testX)
# 数据反归一化
trainPredict = scaler.inverse_transform(trainPredict)
trainY = scaler.inverse_transform([trainY])
testPredict = scaler.inverse_transform(testPredict)
testY = scaler.inverse_transform([testY])
trainScore = math.sqrt(
mean_squared_error(trainY[0], trainPredict[:, 0]))
print('Train Score: %.5f RMSE' % (trainScore))
testScore = math.sqrt(mean_squared_error(testY[0], testPredict[:, 0]))
print('Test Score: %.5f RMSE' % (testScore))
trainPredictPlot = numpy.empty_like(dataset)
trainPredictPlot[:, :] = numpy.nan
trainPredictPlot[self.look_back:len(trainPredict) +
self.look_back, :] = trainPredict
# shift test predictions for plotting
testPredictPlot = numpy.empty_like(dataset)
testPredictPlot[:, :] = numpy.nan
testPredictPlot[len(trainPredict) + (self.look_back * 2) +
1:len(dataset) - 1, :] = testPredict
compare = self.cal_the_return(testPredict, testY)
print('index return={};\nStrategy return={}'.format(
compare['acc_y'].iloc[-1], compare['acc_p0'].iloc[-1]))
# compare[['acc_p0','acc_p1','acc_p2','acc_p3','acc_p4','acc_p5','acc_y']].plot()
plt.figure(figsize=(20, 6))
l1, = plt.plot(compare['acc_y'], color='red', linewidth=5)
l2, = plt.plot(compare['acc_p0'], color='b', linewidth=2)
l3, = plt.plot(compare['acc_p5'], color='g', linewidth=2)
plt.ylabel('Height m')
plt.legend([l1, l2, l3], ('CSI500', 'S0', 'S1'), loc='best')
plt.title('LSTM Prediction--{}'.format(self.model_name))
plt.savefig('./img/收益率曲线_{}.svg'.format(self.model_name), format='svg')
plt.show()