def run_predict(self,imf_list,current_price_index,date,datafile,train_data_dir,train_number,hidden_number,result_number):
decision = 0
power_decision = 0
imf_power_decision = 0
period_decision = 0
extrem_decision = 0
current_rise_flag = 0
data = np.array(list(imf_list[2]))
imf_max_index = argrelextrema(data,np.greater)[0]
imf_min_index = argrelextrema(data,np.less)[0]
price_list = self.close_price[current_price_index-499:current_price_index+1]
price_data = np.array(price_list)
price_max_index = argrelextrema(price_data,np.greater)[0]
print "price_max_index %s"%price_max_index
price_min_index = argrelextrema(price_data,np.less)[0]
print "price_min_index %s"%price_min_index
imf2 = list(imf_list[2])
imf_extrem_decision = 0
if len(list(price_max_index)) > 0 and len(list(price_min_index)) > 0 :
index_list = list(price_max_index)+(list(price_min_index))
print "index list%s"%index_list
index_list.sort()
last_extrem_position = index_list[-2]
distance = 500 - last_extrem_position
true_index = 500-distance
print "distance %s"%distance
print "true index %s"%true_index
data = np.array(list(imf_list[2]))
imf_max_index = argrelextrema(data,np.greater)[0]
print "imf max index %s"%imf_max_index
imf_min_index = argrelextrema(data,np.less)[0]
print "imf min index %s"%imf_min_index
imf_max_true_index = list(filter(lambda n:n<true_index,imf_max_index))
imf_min_true_index = list(filter(lambda n:n<true_index,imf_min_index))
print "imf max true index %s"%imf_max_true_index
print "imf min true index %s"%imf_min_true_index
if list(imf_max_true_index)[-1]>list(imf_min_true_index)[-1]:
buyhigh = imf2[imf_max_true_index[-1]]-imf2[imf_min_true_index[-1]]
buyhigh2 = imf2[imf_max_true_index[-1]]-imf2[-1]
print "buy high %s"%buyhigh
print "buy high2 %s"%buyhigh2
sellhigh = imf2[imf_max_true_index[-1]]-imf2[imf_min_true_index[-1]]
sellhigh2 = imf2[-1]-imf2[imf_min_true_index[-1]]
print "sell high %s"%sellhigh
print "sell high2 %s"%sellhigh2
distance_last_imf_max = 500-list(imf_max_true_index)[-1]
last_min_max = list(imf_max_true_index)[-1] - list(imf_min_true_index)[-1]
cha = distance_last_imf_max- last_min_max
print "cha %s"%(distance_last_imf_max- last_min_max)
# if distance_last_imf_max<10 and distance_last_imf_max>5:
if (buyhigh/buyhigh2)>0.9 and (buyhigh/buyhigh2)<1.1:
imf_extrem_decision = -1
elif list(imf_max_true_index)[-1]<list(imf_min_true_index)[-1]:
buyhigh = imf2[imf_max_true_index[-1]]-imf2[imf_min_true_index[-1]]
buyhigh2 = imf2[imf_max_true_index[-1]]-imf2[-1]
print "buy high %s"%buyhigh
print "buy high2 %s"%buyhigh2
sellhigh = imf2[imf_max_true_index[-1]]-imf2[imf_min_true_index[-1]]
sellhigh2 = imf2[-1]-imf2[imf_min_true_index[-1]]
print "sell high %s"%sellhigh
print "sell high2 %s"%sellhigh2
distance_last_imf_max = 500-list(imf_max_true_index)[-1]
last_min_max = list(imf_max_true_index)[-1] - list(imf_min_true_index)[-1]
cha = distance_last_imf_max- last_min_max
print "cha %s"%(distance_last_imf_max- last_min_max)
# if distance_last_imf_max<10 and distance_last_imf_max>5:
if (sellhigh/sellhigh2)>0.8 and (sellhigh/sellhigh2)<1.2:
imf_extrem_decision = 1
now_position_last_min_delta = len(imf_list[2])-1-imf_min_index[-1]
print "now min delta%s"%now_position_last_min_delta
now_position_last_max_delta = len(imf_list[2])-1-imf_max_index[-1]
print "now max delta%s"%now_position_last_max_delta
print "date %s"%self.date[current_price_index]
print "date %s"%date
print "current price %s"%self.close_price[current_price_index]
print "trade price %s"%self.open_price[current_price_index+1]
print "buy price %s"%self.buy_price
###########################################################################################
recent_price_list = self.close_price[current_price_index-4:current_price_index+1]
recent_price_format = self.format_data(recent_price_list)
recent_price_format2 = recent_price_format[-5:]
print "recent price format %s"%recent_price_format2
############################################################################################
# f = os.popen("tail %s/data_390 -n 2|head -n 1"%train_data_dir)
f = os.popen("cat imf2")
lines = f.readlines()
f.close()
# test_data = line.split(" ")
# test_data.pop()
test_data_float1 = [float(each) for each in lines]
print "raw test data length %s"%len(test_data_float1)
tmp = self.format_data(test_data_float1)
test_data_float = tmp[-60:-10]
print "test data %s"%test_data_float
print "test data length %s"%len(test_data_float)
result = []
fin_result = []
sum = 0
for i in range(1):
neuro_prediction = neuroprediction("%s/data_380"%train_data_dir,"net_file",train_number,hidden_number,0,result_number)
neuro_prediction.train_nn()
result.append(neuro_prediction.test_each(test_data_float))
print result[-1]
for i in range(10):
for j in range(1):
sum += result[j][i]
fin_result.append(sum)
sum = 0
fin_result1 = [fin_result[i]/1 for i in range(10)]
print "recent price format %s"%recent_price_format2
print "predict result %s"%fin_result1
## ######################################################################################
recent_price_format3 = []
for i in recent_price_format2:
if i<0:
beishu1 = (-1)/min(test_data_float)
recent_price_format3.append(i/beishu1)
elif i>0:
beishu2 = 1/max(test_data_float)
recent_price_format3.append(i/beishu2)
else:
recent_price_format3.append(i)
print "recent price format 3 %s"%recent_price_format3
fin_result2 = [(recent_price_format2[i]+fin_result1[i+5])/2 for i in range(len(recent_price_format3))]
print "after add recent price %s"%fin_result2
## #######################################################################################
## #######################################################################################
imf_2 = list(imf_list[2][:-10])
imf = imf_2+fin_result2
print "len imf %s"%len(imf)
print "imf %s"%list(imf_list[2])[-10:]
imf = fin_result1
imf = list(imf_list[2])
imf_ma5_before = (imf[-6]+imf[-5]+imf[-4]+imf[-3]+imf[-2])/5.0
imf_ma3_before = (imf[-4]+imf[-3]+imf[-2])/3.0
imf_ma5_current = (imf[-5]+imf[-4]+imf[-3]+imf[-2]+imf[-1])/5.0
imf_ma3_current = (imf[-3]+imf[-2]+imf[-1])/3.0
imf_power_before = imf[-2] - imf_ma5_before
imf_3_power_before = imf[-2] - imf_ma3_before
imf_power_current = imf[-1] - imf_ma5_current
imf_3_power_current = imf[-1] - imf_ma3_current
print "imf power before %s"%imf_power_before
print "imf power current %s"%imf_power_current
print "last imf power current %s"%self.last_imf_power
print "imf 3 power before %s"%imf_3_power_before
print "imf 3 power current %s"%imf_3_power_current
print "last imf 3 power current %s"%self.last_imf_3_power
last_5 = self.close_price[current_price_index-5:current_price_index]
last_5_mean = np.mean(self.close_price[current_price_index-5:current_price_index])
last_5_std = np.std([last_5[i]-last_5_mean for i in range(5)])
print "std 5 %s"%last_5_std
self.imf_power.append(imf_power_current)
last_10 = self.close_price[current_price_index-6:current_price_index]
####################################################################################################################
imf_decision = 0
data = np.array(last_5)
last_5_max_index = argrelextrema(data,np.greater)[0]
last_5_min_index = argrelextrema(data,np.less)[0]
data = np.array(last_10)
last_10_max_index = argrelextrema(data,np.greater)[0]
last_10_min_index = argrelextrema(data,np.less)[0]
last_10_imf = list(imf_list[2])[-10:]
print "last 10 price %s"%last_10
if len(last_5_max_index)>0 and len(last_5_min_index)>0:
last_10_imf_array = np.array(last_10_imf)
last_10_imf_max_index = argrelextrema(last_10_imf_array,np.greater)[0]
last_10_imf_min_index = argrelextrema(last_10_imf_array,np.less)[0]
if len(last_10_imf_max_index)>0 and last_10_imf_max_index[-1]<4:
imf_decision = -1
if len(last_10_imf_min_index)>0 and last_10_imf_min_index[-1]<4:
imf_decision = 1
#####################################################################################################################
#if (imf_power_before>0 and imf_power_current<0): #or float(imf_power_before)/imf_power_current > 2.5:
#if (imf_3_power_before>0 and imf_3_power_current<0): #or float(imf_power_before)/imf_power_current > 2.5:
#if (self.last_imf_power>0 and imf_power_current<0): #or float(imf_power_before)/imf_power_current > 2.5:
if (imf[-1]<imf[-2]) and ((recent_price_format2[-1]<recent_price_format2[-2])):# and ((recent_price_format2[-1]==1) or (recent_price_format2[-2]==1 and recent_price_format2[-1]!=-1)):# and (recent_price_format2[-1]==1):# or recent_price_format2[-1]==-1): #or float(imf_power_before)/imf_power_current > 2.5:
#if (self.last_imf_3_power>0 and imf_3_power_current<0): #or float(imf_power_before)/imf_power_current > 2.5:
#if (self.last_imf_3_power>imf_3_power_current): #or float(imf_power_before)/imf_power_current > 2.5:
imf_power_decision = 1
#if (imf_power_before<0 and imf_power_current>0):# or float(imf_power_current)/imf_power_before > 2.5:
#if (imf_3_power_before<0 and imf_3_power_current>0):# or float(imf_power_current)/imf_power_before > 2.5:
#if (self.last_imf_3_power<0 and imf_3_power_current>0):# or float(imf_power_current)/imf_power_before > 2.5:
if (imf[-1]>imf[-2]) and ((recent_price_format2[-1]>recent_price_format2[-2]) and (recent_price_format2[-3]>recent_price_format2[-2])):# or (recent_price_format2[-2]==-1 and recent_price_format2[-1]!=1)) :#or ((imf[-1]<imf[-2] and imf[-2]<imf[-3]) and (recent_price_format2[-1]==-1)):# or recent_price_format2[-1]==1) : #or float(imf_power_before)/imf_power_current > 2.5:
#if (self.last_imf_power<0 and imf_power_current>0):# or float(imf_power_current)/imf_power_before > 2.5:
imf_power_decision = -1
print "imf power decision %s"%imf_power_decision
trade_price = self.open_price[current_price_index+1]
current_price = self.close_price[current_price_index]
if self.buy_price!=0 and (imf_extrem_decision == 1) and self.share > 0 :
print "sell 1"
self.money = self.sell(self.share,trade_price)
self.share = 0
self.buy_price = 0
self.total_asset = self.money - self.sell_fee(self.money)
self.last_buy_flag = 0
self.money = self.total_asset
self.sell_x_index.append(current_price_index)
self.sell_y_index.append(current_price)
tomorrow_will = 1
self.down_macd_counter = 0
self.buy_macd = 0
self.holdtime = 0
once_flag = 1
# elif self.buy_price!=0 and ((trade_price-self.buy_price)/self.buy_price) >0.02 and self.share > 0 :
# print "sell 2"
# self.money = self.sell(self.share,trade_price)
# self.share = 0
# self.buy_price = 0
# self.total_asset = self.money - self.sell_fee(self.money)
# self.last_buy_flag = 0
# self.money = self.total_asset
# self.sell_x_index.append(current_price_index)
# self.sell_y_index.append(current_price)
# tomorrow_will = 1
# self.down_macd_counter = 0
# self.buy_macd = 0
# self.holdtime = 0
# once_flag = 1
elif self.money>0 and imf_extrem_decision == -1:
print "buy buy 1"
self.share = self.buy(self.money,trade_price)
self.money = 0
self.buy_price = trade_price
self.total_asset = self.share*trade_price
self.buy_x_index.append(current_price_index)
self.buy_y_index.append(current_price)
self.last_buy_flag = 1
self.buy_macd = self.macd[current_price_index]
tomorrow_will = -1
self.holdtime = 1
else:
self.last_buy_flag = 0
self.last_imf_power = imf_power_current
self.last_imf_3_power = imf_3_power_current
self.total_asset_list.append(self.total_asset)
print "total asset is %s"%(self.total_asset)
print "money is %s"%(self.money)
print "share is %s"%(self.share)
def run_predict(self,imf_list,current_price_index,date,datafile,train_data_dir,train_number,hidden_number,result_number):
decision = 0
power_decision = 0
imf_power_decision = 0
period_decision = 0
extrem_decision = 0
current_rise_flag = 0
data = np.array(imf_list[2])
imf_max_index = argrelextrema(data,np.greater)[0]
imf_min_index = argrelextrema(data,np.less)[0]
now_position_last_min_delta = len(imf_list[2])-1-imf_min_index[-1]
print "now min delta%s"%now_position_last_min_delta
now_position_last_max_delta = len(imf_list[2])-1-imf_max_index[-1]
print "now max delta%s"%now_position_last_max_delta
print "date %s"%self.date[current_price_index]
print "date %s"%date
print "current price %s"%self.close_price[current_price_index]
print "trade price %s"%self.open_price[current_price_index+1]
print "buy price %s"%self.buy_price
###########################################################################################
recent_price_list = self.close_price[current_price_index-4:current_price_index+1]
recent_price_format = self.format_data(recent_price_list)
recent_price_format2 = recent_price_format[-5:]
print "recent price format %s"%recent_price_format2
############################################################################################
# f = os.popen("tail %s/data_390 -n 2|head -n 1"%train_data_dir)
f = os.popen("cat imf2")
lines = f.readlines()
f.close()
# test_data = line.split(" ")
# test_data.pop()
test_data_float1 = [float(each) for each in lines]
print "raw test data length %s"%len(test_data_float1)
tmp = self.format_data(test_data_float1)
test_data_float = tmp[-60:-10]
print "test data %s"%test_data_float
print "test data length %s"%len(test_data_float)
result = []
fin_result = []
sum = 0
for i in range(20):
neuro_prediction = neuroprediction("%s/data_380"%train_data_dir,"net_file",train_number,hidden_number,0,result_number)
neuro_prediction.train_nn()
result.append(neuro_prediction.test_each(test_data_float))
print result[-1]
for i in range(10):
for j in range(20):
sum += result[j][i]
fin_result.append(sum)
sum = 0
fin_result1 = [fin_result[i]/20 for i in range(10)]
print "recent price format %s"%recent_price_format2
print "predict result %s"%fin_result1
## ######################################################################################
recent_price_format3 = []
for i in recent_price_format2:
if i<0:
beishu1 = (-1)/min(test_data_float)
recent_price_format3.append(i/beishu1)
elif i>0:
beishu2 = 1/max(test_data_float)
recent_price_format3.append(i/beishu2)
else:
recent_price_format3.append(i)
print "recent price format 3 %s"%recent_price_format3
fin_result2 = [(recent_price_format3[i]+fin_result1[i+5])/2 for i in range(len(recent_price_format3))]
print "after add recent price %s"%fin_result2
## #######################################################################################
## #######################################################################################
imf_2 = list(imf_list[2][:-10])
imf = imf_2+fin_result2
print "len imf %s"%len(imf)
imf_ma5_before = (imf[-6]+imf[-5]+imf[-4]+imf[-3]+imf[-2])/5.0
imf_ma3_before = (imf[-4]+imf[-3]+imf[-2])/3.0
imf_ma5_current = (imf[-5]+imf[-4]+imf[-3]+imf[-2]+imf[-1])/5.0
imf_ma3_current = (imf[-3]+imf[-2]+imf[-1])/3.0
imf_power_before = imf[-2] - imf_ma5_before
imf_3_power_before = imf[-2] - imf_ma3_before
imf_power_current = imf[-1] - imf_ma5_current
imf_3_power_current = imf[-1] - imf_ma3_current
print "imf power before %s"%imf_power_before
print "imf power current %s"%imf_power_current
print "last imf power current %s"%self.last_imf_power
print "imf 3 power before %s"%imf_3_power_before
print "imf 3 power current %s"%imf_3_power_current
print "last imf 3 power current %s"%self.last_imf_3_power
last_5 = self.close_price[current_price_index-5:current_price_index]
last_5_mean = np.mean(self.close_price[current_price_index-5:current_price_index])
last_5_std = np.std([last_5[i]-last_5_mean for i in range(5)])
print "std 5 %s"%last_5_std
self.imf_power.append(imf_power_current)
#if (imf_power_before>0 and imf_power_current<0): #or float(imf_power_before)/imf_power_current > 2.5:
#if (imf_3_power_before>0 and imf_3_power_current<0): #or float(imf_power_before)/imf_power_current > 2.5:
#if (self.last_imf_power>0 and imf_power_current<0): #or float(imf_power_before)/imf_power_current > 2.5:
if (imf[-1]<imf[-2] and imf[-2]>imf[-3]):# and (recent_price_format2[-1]==1):# or recent_price_format2[-1]==-1): #or float(imf_power_before)/imf_power_current > 2.5:
#if (self.last_imf_3_power>0 and imf_3_power_current<0): #or float(imf_power_before)/imf_power_current > 2.5:
#if (self.last_imf_3_power>imf_3_power_current): #or float(imf_power_before)/imf_power_current > 2.5:
imf_power_decision = 1
#if (imf_power_before<0 and imf_power_current>0):# or float(imf_power_current)/imf_power_before > 2.5:
#if (imf_3_power_before<0 and imf_3_power_current>0):# or float(imf_power_current)/imf_power_before > 2.5:
#if (self.last_imf_3_power<0 and imf_3_power_current>0):# or float(imf_power_current)/imf_power_before > 2.5:
if (imf[-1]>imf[-2] and imf[-2]<imf[-3]) and (recent_price_format2[-1]==-1):# or recent_price_format2[-1]==1) : #or float(imf_power_before)/imf_power_current > 2.5:
#if (self.last_imf_power<0 and imf_power_current>0):# or float(imf_power_current)/imf_power_before > 2.5:
imf_power_decision = -1
print "imf power decision %s"%imf_power_decision
trade_price = self.open_price[current_price_index+1]
current_price = self.close_price[current_price_index]
if self.buy_price!=0 and (imf_power_decision == 1) and self.share > 0 :
print "sell 1"
self.money = self.sell(self.share,trade_price)
self.share = 0
self.buy_price = 0
self.total_asset = self.money - self.sell_fee(self.money)
self.last_buy_flag = 0
self.money = self.total_asset
self.sell_x_index.append(current_price_index)
self.sell_y_index.append(current_price)
tomorrow_will = 1
self.down_macd_counter = 0
self.buy_macd = 0
self.holdtime = 0
once_flag = 1
elif self.money>0 and imf_power_decision == -1:
print "buy buy 1"
self.share = self.buy(self.money,trade_price)
self.money = 0
self.buy_price = trade_price
self.total_asset = self.share*trade_price
self.buy_x_index.append(current_price_index)
self.buy_y_index.append(current_price)
self.last_buy_flag = 1
self.buy_macd = self.macd[current_price_index]
tomorrow_will = -1
self.holdtime = 1
else:
self.last_buy_flag = 0
self.last_imf_power = imf_power_current
self.last_imf_3_power = imf_3_power_current
self.total_asset_list.append(self.total_asset)
print "total asset is %s"%(self.total_asset)
print "money is %s"%(self.money)
print "share is %s"%(self.share)
#print array_after_format
label = list(data_mat[:,-1])
fp=open(train_file,'w')
fp.write(str(row_number))
fp.write(" ")
fp.write(str(col_number-1))
fp.write(" ")
fp.write("1")
fp.write("\n")
for i in range(row_number):
for j in range(col_number-1):
fp.write(str(array_after_format[i][j]))
fp.write(" ")
fp.write("\n")
fp.write(str(label[i]))
fp.write("\n")
fp.close()
my_nn = neuroprediction(train_file,"net_file",4,10,0,1)
my_nn.train_nn()
col_number = data_mat.shape[1]
col_after_format = []
for i in range(col_number - 1):
col_list = list(data_mat[:, i])
#col_after_format.append(format_data(col_list))
col_after_format.append((col_list))
new_array = np.array(col_after_format)
array_after_format = new_array.transpose()
#print array_after_format
label = list(data_mat[:, -1])
fp = open(train_file, 'w')
fp.write(str(row_number))
fp.write(" ")
fp.write(str(col_number - 1))
fp.write(" ")
fp.write("1")
fp.write("\n")
for i in range(row_number):
for j in range(col_number - 1):
fp.write(str(array_after_format[i][j]))
fp.write(" ")
fp.write("\n")
fp.write(str(label[i]))
fp.write("\n")
fp.close()
my_nn = neuroprediction(train_file, "net_file", 4, 10, 0, 1)
my_nn.train_nn()