def allScaler(self):
if self.list_x is not None:
return self.list_x, self.list_y, self.scaler_x, self.scaler_y, self.train_map
df = stock_sql.getStockFrame()
df.sort_values(by=["symbol"], inplace=True, ignore_index=True)
arr = np.array(df)
data_new_x, data_new_y, list_index, self.train_map = self.getDataWithIndexMultiple(
arr, self.only_untrain)
data_new_x = np.concatenate(data_new_x, axis=0)
data_new_y = np.concatenate(data_new_y, axis=0)
self.scaler_x, self.scaler_y = MinMaxScaler(), MinMaxScaler()
data_new_x, data_new_y = self.scaler_x.fit_transform(
data_new_x), self.scaler_y.fit_transform(data_new_y)
self.list_x, self.list_y = {}, {}
list_symbols = list(self.train_map.keys())
for i in range(len(list_index)):
data_a, data_b = data_new_x[:list_index[
i]], data_new_y[:list_index[i]]
data_new_x, data_new_y = data_new_x[list_index[i]:], data_new_y[
list_index[i]:]
self.list_x[list_symbols[i]] = data_a
self.list_y[list_symbols[i]] = data_b
return self.list_x, self.list_y, self.scaler_x, self.scaler_y, self.train_map
def test(self):
print("+++++++++++++++++++++++")
self.verify = True
self.only_untrain = True
df = stock_sql.getStockFrame()
df.sort_values(by=["symbol"], inplace=True, ignore_index=True)
stock_list = np.array(df)
for info in stock_list:
data = self.getTrainData(stockInfo(symbol=info[0]))
data_x, data_y = data[0], data[1]
for i in range(len(data_x)):
print(data_x[i][self.SEQUENCE_LEN - 1])
print(data_y[i])
def dataGenerator(self, index: int = 0):
df = stock_sql.getStockFrame()
df.sort_values(by=["symbol"], inplace=True, ignore_index=True)
# 截取index以后的数据
stock_array = np.array(df)[index:]
batch_array = stock_sql.arr_split(stock_array, self.batch_size)
for s_array in batch_array:
# 多进程
list_x, list_y = self.dataConcat(s_array)
# list_x,list_y =dataConcat(s_array,only_untrain)
if len(list_x[0]) == 0 or len(list_y) == 0: continue
data_x = []
for i in range(self.tuple_x):
data_x.append(np.concatenate(list_x[i], axis=0))
data_y = np.concatenate(list_y, axis=0)
yield data_x, data_y, s_array
def getExVec(self):
df = stock_sql.getStockFrame(
headers=["symbol", "name", "area", "industry"])
df.sort_values(by=["symbol"], inplace=True, ignore_index=True)
df.fillna(" ", inplace=True)
area_vec = OneHotEncoder(sparse=False).fit_transform(df[["area"]])
indu_vec = OneHotEncoder(sparse=False).fit_transform(df[["industry"]])
# data = pd.read_csv("data/stock_basic.csv", dtype=object)
# aa = OneHotEncoder().fit_transform(data[["所在地域"]])
data_area = {}
data_indu = {}
symbols = np.array(df[["symbol"]])
# areas =np.array(df[["area"]]).reshape(-1)
# data_t={}
for i in range(len(symbols)):
data_area[symbols[i][0]] = area_vec[i]
data_indu[symbols[i][0]] = indu_vec[i]
return data_area, data_indu, area_vec.shape, indu_vec.shape
def upDownrate(self):
stock_sql.ClearRecordData()
df = stock_sql.getStockFrame()
stockList = np.array(df)
fitConunt = {}
for i in range(prepare.RREDICT_LEN + 1):
fitConunt[str(i)] = 0
for stock in stockList[0:]:
realData, predictData = self.trainMode.predict(stockInfo(symbol=stock[0]))
if realData.shape[0] != prepare.RREDICT_LEN:
continue
closeColnum = prepare.getTrianColnum(stockInfo(symbol=stock[0]))
real_up, pred_up, count, total = upDownfit(closeColnum, realData, predictData)
print("symbol:" + str(stock[0]) + " name:" + str(stock[1]) + " fit:" + str(count))
fitConunt[str(count)] += 1
x = [i for i in range(prepare.RREDICT_LEN + 1)]
y = [fitConunt[str(d)] for d in range(prepare.RREDICT_LEN + 1)]
plt.plot(x, y)
plt.show()
print(fitConunt)
def upDownrate(self):
df = stock_sql.getStockFrame()
stockList = np.array(df)
fitConunt = {}
for i in range(prepare.RREDICT_LEN + 1):
fitConunt[str(i)] = 0
for stock in stockList[0:]:
realData, predictData = self.trainMode.predict(stockInfo(symbol=stock[0]))
predictData= [np.argmax(predictData[i]) for i in range(len(predictData))]
predictData =np.array(predictData).reshape((-1,1))
if realData.shape[0] != prepare.RREDICT_LEN:
continue
count= self.fitCount(realData,predictData)
print("symbol:" + str(stock[0]) + " name:" + str(stock[1]) + " fit:" + str(count))
fitConunt[str(count)] += 1
x = [i for i in range(prepare.RREDICT_LEN + 1)]
y = [fitConunt[str(d)] for d in range(prepare.RREDICT_LEN + 1)]
plt.plot(x, y)
plt.show()
print(fitConunt)
def RRrate(self):
stock_sql.ClearRecordData()
df = stock_sql.getStockFrame()
stockList = np.array(df)
fitConunt = {}
for stock in stockList[0:]:
realData, predictData = trainning. logisticsTrain().predict(stockInfo(symbol=stock[0]))
if len(realData) is 0:
continue
r2 = r2_socre(realData, predictData, 0)
if r2 in fitConunt.keys():
fitConunt[r2] += 1
else:
fitConunt[r2] = 0
print("symbol:" + str(stock[0]) + " name:" + str(stock[1]) + " fit:" + str(r2))
x = list(fitConunt.keys())
x.sort()
y = [fitConunt[d] for d in x]
x = [math.tanh(d) for d in x]
plt.plot(x, y)
plt.show()
print(fitConunt)
def getList():
print("getList called")
data = stock_sql.getStockFrame()
return data.to_json(orient='records', force_ascii=False)