def wavlet_plt():
# trick to get the axes
fig,ax = plt.subplots()
xValues = close_price[-200:]
ax.plot(xValues, label=code, color="r", linewidth=1)
zValues = cwavelet.getWaveletData(xValues, 'db2', 2, 'sqtwolog')
zxValue = np.arange(0,len(zValues),1)
#plt.figure(figsize=(16,8))
ax.plot(zxValue, zValues, color="b", linewidth=2)
ax.grid()
# make ticks and tick labels
xticks=range(0, len(xValues)+1,10)
#xticklabels=['2000-01-0'+str(n) for n in range(1,len(xValues)+1)]
xticklabels = df['date'].apply(lambda x : str(x)).get_values()
# set ticks and tick labels
ax.set_xticks(xticks)
ax.set_xticklabels(xticklabels,rotation=15)
#plt.legend()
plt.show()
def plotData(stock_code, stock_name):
try:
xValues = []
yValues = []
xLabels = []
dataPath = os.getcwd() + '\\stockdata\\'
i = -1
for line in open(dataPath + stock_code + ".csv"):
f_date, f_open, f_high, f_low, f_close, f_volume, f_adjclose = line.split(
",")
i += 1
if i == 0 or i > 1000:
continue
xValues.append(i)
yValues.append(float(f_adjclose))
xLabels.append(f_date)
yValues.reverse()
zValues = cwavelet.getWaveletData(yValues, 'db2', 4, 'sqtwolog')
zxValue = np.arange(0, len(zValues), 1)
print len(zxValue), len(zValues)
plt.figure(figsize=(16, 8))
plt.plot(xValues, yValues, label=stock_code, color="b", linewidth=1)
plt.plot(zxValue, zValues, color="r", linewidth=2)
plt.xlabel("Time")
plt.ylabel("Price")
plt.title(stock_name)
#plt.xticks(range(min(xLabels), max(xLabels)+1, 10))
plt.grid()
#plt.legend()
plt.show()
except Exception as e:
print("Exception:>>>" + str(e))
finally:
None
def plotData(stock_code, stock_name):
try:
xValues = []
yValues = []
xLabels = []
dataPath = os.getcwd() + "\\stockdata\\"
i = -1
for line in open(dataPath + stock_code + ".csv"):
f_date, f_open, f_high, f_low, f_close, f_volume, f_adjclose = line.split(",")
i += 1
if i == 0 or i > 1000:
continue
xValues.append(i)
yValues.append(float(f_adjclose))
xLabels.append(f_date)
yValues.reverse()
zValues = cwavelet.getWaveletData(yValues, "db2", 4, "sqtwolog")
zxValue = np.arange(0, len(zValues), 1)
print len(zxValue), len(zValues)
plt.figure(figsize=(16, 8))
plt.plot(xValues, yValues, label=stock_code, color="b", linewidth=1)
plt.plot(zxValue, zValues, color="r", linewidth=2)
plt.xlabel("Time")
plt.ylabel("Price")
plt.title(stock_name)
# plt.xticks(range(min(xLabels), max(xLabels)+1, 10))
plt.grid()
# plt.legend()
plt.show()
except Exception as e:
print ("Exception:>>>" + str(e))
finally:
None
def wavlet_plt():
# trick to get the axes
fig, ax = plt.subplots()
xValues = close_price[-200:]
ax.plot(xValues, label=code, color="r", linewidth=1)
zValues = cwavelet.getWaveletData(xValues, 'db2', 2, 'sqtwolog')
zxValue = np.arange(0, len(zValues), 1)
#plt.figure(figsize=(16,8))
ax.plot(zxValue, zValues, color="b", linewidth=2)
ax.grid()
# make ticks and tick labels
xticks = range(0, len(xValues) + 1, 10)
#xticklabels=['2000-01-0'+str(n) for n in range(1,len(xValues)+1)]
xticklabels = df['date'].apply(lambda x: str(x)).get_values()
# set ticks and tick labels
ax.set_xticks(xticks)
ax.set_xticklabels(xticklabels, rotation=15)
#plt.legend()
plt.show()
def plotRateOfReturn(stock_code):
try:
xValues = []
yValues = []
i = -1
for line in open(stock_code + ".csv"):
i += 1
if i == 0 or i > 1000:
continue
f_date, f_open, f_high, f_low, f_close, f_volume, f_adjclose = line.split(",")
yValues.append(float(f_adjclose))
# yValues删除最后一个元素，zValues删除第一个元素
zValues = copy.deepcopy(yValues) # 深拷贝
yValues.reverse()
yValues.pop()
zValues.pop()
zValues.reverse()
if len(yValues) != len(zValues):
return
rateValues = []
for i in range(0, len(yValues)):
print float(zValues[i]) / yValues[i]
rateValues.append(math.log(float(zValues[i]) / yValues[i]))
xValues.append(i)
rateValues = cwavelet.getWaveletData(yValues, "db4", 2, "sqtwolog")
# BP神经网络
# patStock = []
# for i in range(0, len(yValues)):
# each = [[i], [yValues[i]]]
# patStock.append(each)
# patStockPre = copy.deepcopy(patStock)
# for i in range(len(yValues), len(yValues)+10):
# each = [[i], [0]]
# patStockPre.append(each)
# pat = [
# [[0], [0]],
# [[2], [1]],
# [[3], [1]],
# [[4], [5]]
# ]
#
# # create a network with two input, two hidden, and one output nodes
# n = BP.NN(1, 2, 1)
# # train it with some patterns
# n.train(patStock)
# # test it
# n.test(patStock)
# 最小二乘法
print "原始长度:", len(rateValues)
catRateValues = rateValues[:-7]
print "原始长度:", len(catRateValues)
leastsqValues = cwavelet.getWavePacketData(catRateValues, "haar", 4, 3)
# leastsqValues = cleastsq.getFitYValues(range(len(catRateValues)), catRateValues, range(len(catRateValues)+3))
print "变换后长度:", len(leastsqValues)
newLeastsqValues = np.concatenate((rateValues[:-3], leastsqValues[-3:]))
newLeastsqValues2 = []
for data in newLeastsqValues:
data -= 0.2
newLeastsqValues2.append(data)
newLeastsqValues = newLeastsqValues2
print "变换后长度:", len(newLeastsqValues)
plt.figure(figsize=(16, 8))
plt.legend()
plt.plot(range(len(rateValues)), rateValues, "b-", linewidth=1)
plt.plot(range(len(newLeastsqValues)), newLeastsqValues, "r-", linewidth=1)
plt.xlabel("Time")
plt.ylabel("Price")
plt.title(stock_code)
plt.grid()
plt.show()
except Exception as e:
print ("Exception:>>>" + str(e))
finally:
None
def plotRateOfReturn(stock_code):
try:
xValues = []
yValues = []
i = -1
for line in open(stock_code + ".csv"):
i += 1
if i == 0 or i > 1000:
continue
f_date, f_open, f_high, f_low, f_close, f_volume, f_adjclose = line.split(
",")
yValues.append(float(f_adjclose))
#yValues删除最后一个元素，zValues删除第一个元素
zValues = copy.deepcopy(yValues) #深拷贝
yValues.reverse()
yValues.pop()
zValues.pop()
zValues.reverse()
if len(yValues) != len(zValues):
return
rateValues = []
for i in range(0, len(yValues)):
print float(zValues[i]) / yValues[i]
rateValues.append(math.log(float(zValues[i]) / yValues[i]))
xValues.append(i)
rateValues = cwavelet.getWaveletData(yValues, 'db4', 2, 'sqtwolog')
# BP神经网络
# patStock = []
# for i in range(0, len(yValues)):
# each = [[i], [yValues[i]]]
# patStock.append(each)
# patStockPre = copy.deepcopy(patStock)
# for i in range(len(yValues), len(yValues)+10):
# each = [[i], [0]]
# patStockPre.append(each)
# pat = [
# [[0], [0]],
# [[2], [1]],
# [[3], [1]],
# [[4], [5]]
# ]
#
# # create a network with two input, two hidden, and one output nodes
# n = BP.NN(1, 2, 1)
# # train it with some patterns
# n.train(patStock)
# # test it
# n.test(patStock)
#最小二乘法
print "原始长度:", len(rateValues)
catRateValues = rateValues[:-7]
print "原始长度:", len(catRateValues)
leastsqValues = cwavelet.getWavePacketData(catRateValues, 'haar', 4, 3)
#leastsqValues = cleastsq.getFitYValues(range(len(catRateValues)), catRateValues, range(len(catRateValues)+3))
print "变换后长度:", len(leastsqValues)
newLeastsqValues = np.concatenate(
(rateValues[:-3], leastsqValues[-3:]))
newLeastsqValues2 = []
for data in newLeastsqValues:
data -= 0.2
newLeastsqValues2.append(data)
newLeastsqValues = newLeastsqValues2
print "变换后长度:", len(newLeastsqValues)
plt.figure(figsize=(16, 8))
plt.legend()
plt.plot(range(len(rateValues)), rateValues, 'b-', linewidth=1)
plt.plot(range(len(newLeastsqValues)),
newLeastsqValues,
'r-',
linewidth=1)
plt.xlabel('Time')
plt.ylabel('Price')
plt.title(stock_code)
plt.grid()
plt.show()
except Exception as e:
print("Exception:>>>" + str(e))
finally:
None
def tread_track_backtest(code, df=None):
df = get_stock_k_line(code)
# 分析某个时间段的股票
#dateS = datetime.datetime.today().date() + datetime.timedelta(-100)
#date_start = dateS.strftime("%Y-%m-%d")
#df = df[df.index > date_start]
#print df
# try:
# df = df.reindex(df.index[::-1])
# except Exception as e:
# print str(e)
# return
#print df
close_price = df['close'].get_values()
ma_short = pd.rolling_mean(df['close'], prama_ma_short)
ma_long = pd.rolling_mean(df['close'], prama_ma_long)
#print ma_short
signal = SIGNAL_SALE
print '交易开始'
plt.figure(figsize=(16,8))
#plt.plot(range(len(ma_short)), ma_short.get_values(), label=code, color="b", linewidth=1)
plt.xlabel("Time")
plt.ylabel("Price")
# 判断极点
# for i in range(prama_ma_short+1, len(ma_short)-1):
# if ma_short[i] > ma_short[i-1] and ma_short[i] > ma_short[i+1]:
# print '极大值:', ma_short[i], 'pos:', i
#
# elif ma_short[i] < ma_short[i-1] and ma_short[i] < ma_short[i+1]:
# print '极小值:', ma_short[i], 'pos:', i
# 过滤微小波动
extreIndex = -1 #极值点的索引
for i in range(prama_ma_short+1, len(ma_short)-1):
bMax = ma_short[i] > ma_short[i-1] and ma_short[i] > ma_short[i+1] # 极大值的条件
bMin = ma_short[i] < ma_short[i-1] and ma_short[i] < ma_short[i+1] # 极小值的条件
if bMax or bMin:
extreIndex = i
elif extreIndex > 0:
if ma_short[i] > ma_short[extreIndex]*(1-filter_range) and \
ma_short[i] < ma_short[extreIndex]*(1+filter_range):
ma_short[i] = ma_short[extreIndex]
#print i,ma_short[i]
#plt.plot(df.index, ma_short.get_values(), label=code, color="r", linewidth=1)
# trick to get the axes
fig,ax = plt.subplots()
xValues = close_price[-200:]
ax.plot(xValues, label=code, color="r", linewidth=1)
zValues = cwavelet.getWaveletData(xValues, 'db2', 2, 'sqtwolog')
zxValue = np.arange(0,len(zValues),1)
#plt.figure(figsize=(16,8))
ax.plot(zxValue, zValues, color="b", linewidth=2)
ax.grid()
# make ticks and tick labels
xticks=range(0, len(xValues)+1,10)
xticklabels=['2000-01-0'+str(n) for n in range(1,len(xValues)+1)]
# set ticks and tick labels
ax.set_xticks(xticks)
ax.set_xticklabels(xticklabels,rotation=15)
#plt.legend()
plt.show()
# 交易次数 count_sale = 0
count_buy = 0
min_index_pre = 0 #前一个极小值
max_index_pre = 0 #前一个极大值
# 止损位
keep_stop_price = 0
keep_stop_index = 0
# 止赢位
keep_win_price = 0
keep_win_index = 0
total = 0
price_buy = 0
price_init = 0
money_init = 50000
stock_money = money_init
stock_count = 0
#for i in range(prama_ma_short+1, len(ma_short)-1):
for i in range(len(ma_short)-180, len(ma_short)-1):
#滤波后的均线走平时（即处于滤波区间内），将其识别为一个点
index_post = i+1
try:
while(ma_short[index_post] == ma_short[i] and index_post < len(ma_short)-1):
index_post += 1
except Exception as e:
print str(e)
# 长均线保护策略
bLongMA_protect_close = True
try:
bLongMA_protect_close = ma_short[i] > ma_long[i] # 长均线保护是否关闭
except Exception as e:
print str(e)
#if bLongMA_protect_close == False:
#print "长均线保护打开", i
# 高低点比较策略
if ma_short[i] > ma_short[i-1] and ma_short[i] > ma_short[index_post]:
#print '极大值:', ma_short[i], 'pos:', i
if bLongMA_protect_close and max_index_pre > 0 and ma_short[i] < ma_short[max_index_pre] + drift*(i-max_index_pre) and signal == SIGNAL_BUY:
signal = SIGNAL_SALE
print '卖出:', close_price[i], 'pos:', i
#count_sale += 1
total += close_price[i] - price_buy
stock_money += stock_count * close_price[i]
stock_count = 0
max_index_pre = i
elif ma_short[i] < ma_short[i-1] and ma_short[i] < ma_short[index_post]:
#print '极小值:', ma_short[i], 'pos:', i
if bLongMA_protect_close and min_index_pre > 0 and ma_short[i] > ma_short[min_index_pre] + drift*(i-min_index_pre) and signal == SIGNAL_SALE:
signal = SIGNAL_BUY
print '买入:', close_price[i], 'pos:', i
count_buy += 1
price_buy = close_price[i]
if price_init == 0:
price_init = price_buy
stock_count = (stock_money/100)/close_price[i]*100
stock_money = stock_money - stock_count*close_price[i]
min_index_pre = i
# 高低点突破策略
# 股价突破前一个低点加漂移项，则卖出
elif bLongMA_protect_close and signal == SIGNAL_BUY and min_index_pre > 0 and \
ma_short[i] < ma_short[min_index_pre] + drift*(i-min_index_pre):
signal = SIGNAL_SALE
print '卖出:', close_price[i], 'pos:', i
#count_sale += 1
total += close_price[i] - price_buy
stock_money += stock_count * close_price[i]
stock_count = 0
# 股价突破前一个高点加漂移项，则买入
elif bLongMA_protect_close and signal == SIGNAL_SALE and max_index_pre > 0 and \
ma_short[i] > ma_short[max_index_pre] + drift*(i-max_index_pre):
signal = SIGNAL_BUY
print '买入:', close_price[i], 'pos:', i
count_buy += 1
price_buy = close_price[i]
stock_count = (stock_money/100)/close_price[i]*100
stock_money = stock_money - stock_count*close_price[i]
# 大波段保护策略
elif min_index_pre > 0 and ma_short[i] >= ma_short[min_index_pre]*(1+param_big_band):
keep_stop_price = ma_short[i] * (1-protect_big_band) #止损位
keep_stop_index = i
elif bLongMA_protect_close and signal == SIGNAL_BUY and keep_stop_price > 0 and \
ma_short[i] < keep_stop_price + drift*(i-keep_stop_index):
signal = SIGNAL_SALE
print '卖出:', close_price[i], 'pos:', i
#count_sale += 1
total += close_price[i] - price_buy
stock_money += stock_count * close_price[i]
stock_count = 0
print "buy count:", count_buy
#print "sale count:", count_sale
if stock_count > 0:
stock_money += stock_count * close_price[-1]
total += close_price[-1] - price_buy
print "每股盈利：", total, "收益率：", (stock_money-money_init)/money_init*100,"%\n"
print '交易结束'
plt.grid()
#plt.legend()
#plt.show()
return (stock_money-money_init)/money_init*100
