def get_emotion_hour(start, end, out_name):
'''
:param start:
:param end:
:param out_name:
:return:
'''
out_file = open(out_name, 'w')
dt = start
coll = mongo_handler.get_stock_coll()
title = ['date', 'anger', 'disgust', 'joy', 'sadness', 'fear', 'sum']
out_file.write(','.join(title) + '\n')
while dt <= end:
print(dt)
docs = get_mood_hour(coll, dt)
hour = 0
for doc in docs:
x = ['{}-{}-{} {}'.format(dt[:4], dt[4:6], dt[6:], hour)]
mood = doc['mood']
try:
m_sum = doc['stat']['m_sum']
except:
m_sum = float(mood['0']) + float(mood['1']) + float(
mood['2']) + float(mood['3']) + float(mood['4'])
# 绝对情绪
x.append(str(float(mood['0'])))
x.append(str(float(mood['1'])))
x.append(str(float(mood['2'])))
x.append(str(float(mood['3'])))
x.append(str(float(mood['4'])))
x.append(str(float(m_sum)))
out_file.write(','.join(x) + '\n')
hour += 1
dt = time_tool.add_day_rule(dt, '%Y%m%d')
def get_mood_series():
coll = mongo_handler.get_stock_coll()
dt = '20141201'
end = '20150916'
while dt <= end:
di_mood = get_mood_day(coll, dt)
print(dt, di_mood['mood']['0'], di_mood['mood']['1'], di_mood['mood']['2'],\
di_mood['mood']['3'], di_mood['mood']['4'], di_mood['stat']['m_sum'])
dt = time_tool.add_day_rule(dt, '%Y%m%d')
def write_mood(start, end, out_name):
coll = mongo_handler.get_stock_coll()
out = open(out_name, 'w')
dt = start
while dt < end:
print(dt)
mood = get_mood_day(coll, dt)['mood']
out.write(dt + ' ' + str(mood['0']) + ' ' + str(mood['1']) + ' ' +
str(mood['2']) + ' ' + str(mood['3']) + ' ' +
str(mood['4']) + '\n')
dt = time_tool.add_day_rule(dt, '%Y%m%d')
out.close()
def get_big_board_mood(fi, fi_train):
'''
获取涨跌率及情绪，收盘后及开盘前
'''
file_train = open(fi_train, 'w')
coll = mongo_handler.get_stock_coll()
for line in open(fi):
dt, extent = line.strip().split(',')
dt = datetime.strptime(dt, '%Y/%m/%d').strftime('%Y%m%d')
data = get_mood_close_open(coll, dt)
if data:
print(data)
x = deal_with_x(data)
y = extent
# y = '1' if float(extent) >= 0.0 else '0'
print(y)
file_train.write(y + '\t' + '\t'.join([str(xi)
for xi in x]) + '\n')
def get_professional_mood(fi, fi_train):
'''
获取涨跌率及情绪，专家经验只使用2、4数据，输出结果为拟合所需，前一天的总量
'''
file_train = open(fi_train, 'w')
coll = mongo_handler.get_stock_coll()
for line in open(fi):
dt, extent = line.strip().split(',')
dt = datetime.datetime.strptime(dt, '%Y/%m/%d').strftime('%Y%m%d')
print('训练数据获取中：', dt)
dt = dec_day(dt, 1) # 获取前一天的日期
data = get_mood_day(coll, dt)
if data:
mood = data['mood']
# m_sum = data['stat']['m_sum']
# 分类问题
# y = '1' if float(extent) >= 0.0 else '0'
# x = list()
# x.append(str(float(mood['2'])))
# x.append(str(float(mood['4'])))
# # file_train.write(y + '\t' + '\t'.join(x) + '\n')
# file_train.write(y + ' ' + ' '.join([str(i+1) + ':' + str(xi) for i, xi in enumerate(x)]) + '\n')
# 分类问题
# y = '1' if float(extent) >= 0.0 else '0'
#if float(extent) >= 2.0:
# y = '1'
# elif float(extent) <= -2.0:
# y = '-1'
# else:
# y = '0'
y = extent
x = list()
x.append(str(float(mood['0']) / 1000))
x.append(str(float(mood['1']) / 1000))
x.append(str(float(mood['2']) / 1000))
x.append(str(float(mood['3']) / 1000))
x.append(str(float(mood['4']) / 1000))
# file_train.write(y + '\t' + '\t'.join(x) + '\n')
file_train.write(y + ' ' + ' '.join(x) + '\n')
def get_emotion(start, end, out_name):
'''
获取情绪数据, 按天
:param start:
:param end:
:param out_name:
:return:
'''
out_file = open(out_name, 'w')
dt = start
coll = mongo_handler.get_stock_coll()
title = ['date', 'anger', 'disgust', 'joy', 'sadness', 'fear', 'volume']
out_file.write(','.join(title) + '\n')
while dt <= end:
print(dt)
x = [dt[:4] + '-' + dt[4:6] + '-' + dt[6:]]
# x = [dt[4:6] + '-' + dt[6:]]
data = get_mood_day(coll, dt)
try:
mood = data['mood']
except:
print('ERROR:', "mood = data['mood']")
print(data)
break
try:
amount = data['stat']['m_sum']
except:
amount = float(mood['0']) + float(mood['1']) + float(
mood['2']) + float(mood['3']) + float(mood['4'])
x.append(str(float(mood['0'])))
x.append(str(float(mood['1'])))
x.append(str(float(mood['2'])))
x.append(str(float(mood['3'])))
x.append(str(float(mood['4'])))
x.append(str(float(amount)))
out_file.write(','.join(x) + '\n')
dt = time_tool.add_day_rule(dt, '%Y%m%d')
out_file.close()
def get_big_board_mood_old(fi, fi_train):
'''
获取涨跌率及情绪，已经弃用
'''
file_train = open(fi_train, 'w')
coll = mongo_handler.get_stock_coll()
for line in open(fi):
dt, extent = line.strip().split(',')
dt = datetime.datetime.strptime(dt, '%Y/%m/%d').strftime('%Y%m%d')
dt = dec_day(dt, 1)
data = get_mood_day(coll, dt)
if data:
mood = data['mood']
m_sum = data['stat']['m_sum']
y = extent
x = list()
x.append(str(float(mood['0']) / m_sum * 10))
x.append(str(float(mood['1']) / m_sum * 10))
x.append(str(float(mood['2']) / m_sum * 10))
x.append(str(float(mood['3']) / m_sum * 10))
x.append(str(float(mood['4']) / m_sum * 10))
x.append(str(float(m_sum) / 30000))
file_train.write(y + '\t' + '\t'.join(x) + '\n')
def get_train_data_hour(start, end, out_name):
'''
获取收盘数据，前n天的各个情绪数据，真实交易数据是由tushare获取
'''
# 获取真实大盘数据
results = ts.get_hist_data('sh', start=start, end=end)
print('成功获取 -> 历史大盘数据 ... ...')
# 取得时间序列
series = list(results.index)
print(series)
# 获取数据库连接
coll = mongo_handler.get_stock_coll()
# 遍历每天交易实际数据
for i, date in enumerate(series):
# 第一天留出
if i == 0: continue
# 日期及真实数据
today = datetime.datetime.strptime(date, '%Y-%m-%d').strftime('%Y%m%d')
price_close = results.loc[date]['close']
price_open = results.loc[date]['open']
price_high = results.loc[date]['high']
price_low = results.loc[date]['low']
volume = results.loc[date]['volume'] * 100
last_close = results.loc[series[i - 1]]['close']
extent = results.loc[date]['p_change']
# 涨跌幅当日最大最小
extent_open = (price_open - last_close) / last_close * 100
extent_high = (price_high - last_close) / last_close * 100
extent_low = (price_low - last_close) / last_close * 100
# 创造目标函数
y = list()
# y.append(today)
y.append(date)
# 收盘价作为预测目标
y.append(str(price_close))
# 涨跌幅作为预测目标
# if abs(extent) > 5.0: continue
# y.append(str(extent))
# y.append(str(price_close))
# y.append('1' if extent > 0 else '0')
# y = '1' if extent > 0 else '0'
# 报告中的公式作为预测目标
# y = str(math.log(price_close) - math.log(price_open))
# 其它标准
# y.append(str(extent_high))
# y.append(str(extent_low))
# y.append('1' if extent_open > 0 else '0')
y.append('%.4f' % (extent))
y.append('%.4f' % (extent_open))
y.append('%.4f' % (extent_high))
y.append('%.4f' % (extent_low))
y.append(str(volume))
def get_last_mood(dt, n):
'''
内嵌函数
获取前n天的情绪数据
'''
print(dt, n)
dt_n = dec_day(dt, n)
return get_mood_day(coll, dt_n)
x = []
print('训练数据获取中：', today)
# for i in range(1, 21): # 获取前n天的数据，并且进行预处理
for i in range(0, 1): # 格兰杰, 取当天数据
# 获取训练数据日期
train_date = dec_trade_day(today, series, i)
if train_date == False: break
print('训练数据日期：', train_date)
try:
# 一天数据
docs = get_mood_hour(coll, train_date)
except TypeError:
break
for doc in docs:
mood = doc['mood']
m_sum = doc['stat']['m_sum']
# 绝对情绪
x.append(str(float(mood['0'])))
x.append(str(float(mood['1'])))
x.append(str(float(mood['2'])))
x.append(str(float(mood['3'])))
x.append(str(float(mood['4'])))
x.append(str(float(m_sum)))
# five_sum = float(mood['0']) + float(mood['1']) + float(mood['2']) + float(mood['3']) + float(mood['4'])
# x.append(str(five_sum))
# 绝对情绪修正
# x.append(str(float(mood['0']) / 1000))
# x.append(str(float(mood['1']) / 1000))
# x.append(str(float(mood['2']) / 1000))
# x.append(str(float(mood['3']) / 1000))
# x.append(str(float(mood['4']) / 1000))
# 绝对情绪，取Log
# x.append(str(math.log(float(mood['0']))))
# x.append(str(math.log(float(mood['1']))))
# x.append(str(math.log(float(mood['2']))))
# x.append(str(math.log(float(mood['3']))))
# x.append(str(math.log(float(mood['4']))))
# 数据总和及情绪比例
# x.append(str(float(m_sum) / 100000))
# 总量
# five_sum = float(mood['0']) + float(mood['1']) + float(mood['2']) + float(mood['3']) + float(mood['4'])
# x.append(str(float(mood['0']) / five_sum))
# x.append(str(float(mood['1']) / five_sum))
# x.append(str(float(mood['2']) / five_sum))
# x.append(str(float(mood['3']) / five_sum))
# x.append(str(float(mood['4']) / five_sum))
# 保留小数点后两位
# x.append('%.4f' % (float(mood['0']) / five_sum))
# x.append('%.4f' % (float(mood['1']) / five_sum))
# x.append('%.4f' % (float(mood['2']) / five_sum))
# x.append('%.4f' % (float(mood['3']) / five_sum))
# x.append('%.4f' % (float(mood['4']) / five_sum))
# x.append(str(float(mood['0']) / float(m_sum)))
# x.append(str(float(mood['1']) / float(m_sum)))
# x.append(str(float(mood['2']) / float(m_sum)))
# x.append(str(float(mood['3']) / float(m_sum)))
# x.append(str(float(mood['4']) / float(m_sum)))
# 各种组合情绪，差值
# x.append( str(float(mood['2']) - float(mood['0'])) )
# x.append( str(float(mood['2']) - float(mood['1'])) )
# x.append( str(float(mood['2']) - float(mood['3'])) )
# x.append( str(float(mood['2']) - float(mood['4'])) )
# 比值
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['0'])))) )
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['1'])))) )
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['2'])))) )
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['3'])))) )
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['4'])))) )
# 成交量与成交金额，取Log
# x.append(str(math.log(amount)))
# x.append(str(math.log(sum_money)))
else:
# 将训练数据写入到文件
file_train = open(out_name, 'a')
file_train.write(','.join(y) + ',' + ','.join(x) + '\n')
print(','.join(y) + ',' + ','.join(x))
def get_real_data_file(fi, fi_train):
'''
获取收盘数据，及当天情绪
'''
file_train = open(fi_train, 'w')
coll = mongo_handler.get_stock_coll()
is_instruction = True
# 取得时间序列
series = get_date_series(fi)
for line in open(fi):
if is_instruction:
is_instruction = False
continue
print(line)
real_big_board = line.strip().split(',')
# 日期
# dt = datetime.datetime.strptime(real_big_board[0], '%Y/%m/%d').strftime('%Y%m%d')
# 日期及真实数据
# today = real_big_board[0]
today = datetime.datetime.strptime(real_big_board[0],
'%Y/%m/%d').strftime('%Y%m%d')
price_close = float(real_big_board[1])
price_high = float(real_big_board[2])
price_low = float(real_big_board[3])
price_open = float(real_big_board[4])
extent_yesterday = float(real_big_board[5])
extent = float(real_big_board[7])
amount = float(real_big_board[8])
sum_money = float(real_big_board[9])
# 涨跌幅当日最大最小
extent_high = (price_high - extent_yesterday) / extent_yesterday * 100
extent_low = (price_low - extent_yesterday) / extent_yesterday * 100
print('训练数据获取中：', today)
y = list()
y = [real_big_board[0]]
# y.append(str(extent))
# y.append(str(extent_high))
# y.append(str(extent_low))
x = list()
doc = get_mood_day(coll, today)
if not doc:
print(today, doc)
continue
mood = doc['mood']
m_sum = doc['stat']['m_sum']
# 绝对情绪
x.append(str(float(mood['0'])))
x.append(str(float(mood['1'])))
x.append(str(float(mood['2'])))
x.append(str(float(mood['3'])))
x.append(str(float(mood['4'])))
x.append(str(m_sum))
# file_train.write(' '.join(x) + '\n')
file_train.write('\t'.join(y) + '\t' + '\t'.join(x) + '\n')
def get_train_data_from_csv(fi, fi_train):
'''ls
获取收盘数据，前n天的各个情绪数据
'''
file_train = open(fi_train, 'w')
coll = mongo_handler.get_stock_coll()
is_instruction = True
# 取得时间序列
series = get_date_series(fi)
for line in open(fi):
if is_instruction:
is_instruction = False
continue
print(line)
real_big_board = line.strip().split(',')
# 日期
# dt = datetime.datetime.strptime(real_big_board[0], '%Y/%m/%d').strftime('%Y%m%d')
# 日期及真实数据
today = real_big_board[0]
price_close = float(real_big_board[1])
price_high = float(real_big_board[2])
price_low = float(real_big_board[3])
price_open = float(real_big_board[4])
extent_yesterday = float(real_big_board[5])
extent = float(real_big_board[7])
amount = float(real_big_board[8])
sum_money = float(real_big_board[9])
# 涨跌幅当日最大最小
extent_open = (price_open - extent_yesterday) / extent_yesterday * 100
extent_high = (price_high - extent_yesterday) / extent_yesterday * 100
extent_low = (price_low - extent_yesterday) / extent_yesterday * 100
print('训练数据获取中：', today)
def get_last_mood(dt, n):
'''
内嵌函数
获取前n天的情绪数据
'''
print(dt, n)
dt_n = dec_day(dt, n)
return get_mood_day(coll, dt_n)
y = list()
# y.append(price_close)
# y.append(price_open)
# y.append(price_high)
# y.append(price_low)
# y.append(today)
# 收盘价作为预测目标
# y = str(price_close)
# 涨跌幅作为预测目标
# if abs(extent) > 5.0: continue
# y.append(str(extent))
y.append('1' if extent > 0 else '0')
# y = '1' if extent > 0 else '0'
# 报告中的公式作为预测目标
# y = str(math.log(price_close) - math.log(price_open))
# 其它标准
# y.append(str(extent_high))
# y.append(str(extent_low))
y.append('1' if extent_open > 0 else '0')
# y.append('%.4f' % (extent_open))
y.append('%.4f' % (extent_high))
y.append('%.4f' % (extent_low))
y.append(str(amount))
x = list()
for i in range(1, 6): # 获取前n天的数据，并且进行预处理
# for i in range(0, 1): # 格兰杰
# 获取训练数据日期
train_date = dec_trade_day(today, series, i)
if train_date == False:
break
print('训练数据日期：', train_date)
try:
# 一天数据
doc = get_mood_day(coll, train_date)
# 交易时段的数据
# doc = get_trade_mood(coll, train_date)
mood = doc['mood']
m_sum = doc['stat']['m_sum']
except TypeError:
break
# 绝对情绪
# x.append(str(float(mood['0']) / 4800))
# x.append(str(float(mood['1']) / 4900))
# x.append(str(float(mood['2']) / 9700))
# x.append(str(float(mood['3']) / 2500))
# x.append(str(float(mood['4']) / 10200))
# x.append(str(m_sum))
# 绝对情绪修正
# x.append(str(float(mood['0']) / 1000))
# x.append(str(float(mood['1']) / 1000))
# x.append(str(float(mood['2']) / 1000))
# x.append(str(float(mood['3']) / 1000))
# x.append(str(float(mood['4']) / 1000))
# 绝对情绪，取Log
# x.append(str(math.log(float(mood['0']))))
# x.append(str(math.log(float(mood['1']))))
# x.append(str(math.log(float(mood['2']))))
# x.append(str(math.log(float(mood['3']))))
# x.append(str(math.log(float(mood['4']))))
# 数据总和及情绪比例
# x.append(str(float(m_sum) / 100000))
# 总量
# five_sum = float(mood['0']) + float(mood['1']) + float(mood['2']) + float(mood['3']) + float(mood['4'])
# x.append(str(float(mood['0']) / five_sum))
# x.append(str(float(mood['1']) / five_sum))
# x.append(str(float(mood['2']) / five_sum))
# x.append(str(float(mood['3']) / five_sum))
# x.append(str(float(mood['4']) / five_sum))
# 保留小数点后两位
# x.append('%.4f' % (float(mood['0']) / five_sum))
# x.append('%.4f' % (float(mood['1']) / five_sum))
# x.append('%.4f' % (float(mood['2']) / five_sum))
# x.append('%.4f' % (float(mood['3']) / five_sum))
# x.append('%.4f' % (float(mood['4']) / five_sum))
# x.append(str(float(mood['0']) / float(m_sum)))
# x.append(str(float(mood['1']) / float(m_sum)))
# x.append(str(float(mood['2']) / float(m_sum)))
# x.append(str(float(mood['3']) / float(m_sum)))
# x.append(str(float(mood['4']) / float(m_sum)))
# 各种组合情绪，差值
# x.append( str(float(mood['2']) - float(mood['0'])) )
# x.append( str(float(mood['2']) - float(mood['1'])) )
# x.append( str(float(mood['2']) - float(mood['3'])) )
# x.append( str(float(mood['2']) - float(mood['4'])) )
# 比值
x.append(
str(math.log((1 + float(mood['2'])) / (1 + float(mood['0'])))))
x.append(
str(math.log((1 + float(mood['2'])) / (1 + float(mood['1'])))))
x.append(
str(math.log((1 + float(mood['2'])) / (1 + float(mood['2'])))))
x.append(
str(math.log((1 + float(mood['2'])) / (1 + float(mood['3'])))))
x.append(
str(math.log((1 + float(mood['2'])) / (1 + float(mood['4'])))))
# 成交量与成交金额，取Log
# x.append(str(math.log(amount)))
# x.append(str(math.log(sum_money)))
else:
# 将训练数据写入到文件
file_train.write(' '.join(y) + ' ' + ' '.join(x) + '\n')
def get_mood_last_n(fi, fi_train):
'''
获取涨跌率及情绪，前n天的各个情绪数据
'''
file_train = open(fi_train, 'w')
coll = mongo_handler.get_stock_coll()
is_instruction = True
for line in open(fi):
if is_instruction:
is_instruction = False
continue
print(line)
real_big_board = line.strip().split(',')
# 日期
dt = datetime.datetime.strptime(real_big_board[0],
'%Y/%m/%d').strftime('%Y%m%d')
price_close = float(real_big_board[1])
price_high = float(real_big_board[2])
price_low = float(real_big_board[3])
price_open = float(real_big_board[4])
extent = float(real_big_board[7])
print('训练数据获取中：', dt)
def get_last_mood(dt, n):
'''
内嵌函数
获取前n天的情绪数据
'''
dt_n = dec_day(dt, n)
return get_mood_day(coll, dt_n)
# 涨跌幅作为预测目标
# if abs(extent) > 5.0: continue
y = str(extent)
# 报告中的公式作为预测目标
# y = str(math.log(price_close) - math.log(price_open))
x = list()
# 获取前n(7)天的数据，并且进行预处理
for i in range(1, 2):
# mood = get_last_mood(dt, i)['mood']
# granger revise
mood = get_mood_day(coll, dt)['mood']
m_sum = get_last_mood(dt, i)['stat']['m_sum']
# 情绪注释
# print(count, i, '情绪0'; count += 1
# print(count, i, '情绪1'; count += 1
# print(count, i, '情绪2'; count += 1
# print(count, i, '情绪3'; count += 1
# print(count, i, '情绪4'; count += 1
# print(count, i, '总量'; count += 1
# print(count, i, '比例0'; count += 1
# print(count, i, '比例1'; count += 1
# print(count, i, '比例2'; count += 1
# print(count, i, '比例3'; count += 1
# print(count, i, '比例4'; count += 1
# 绝对情绪
x.append(str(float(mood['0']) / 10000))
x.append(str(float(mood['1']) / 10000))
x.append(str(float(mood['2']) / 10000))
x.append(str(float(mood['3']) / 10000))
x.append(str(float(mood['4']) / 10000))
# 取Log
# x.append(str(math.log(float(mood['0']))))
# x.append(str(math.log(float(mood['1']))))
# x.append(str(math.log(float(mood['2']))))
# x.append(str(math.log(float(mood['3']))))
# x.append(str(math.log(float(mood['4']))))
# 数据总和及情绪比例
# x.append(str(float(m_sum) / 10000))
# x.append(str(float(mood['0']) / float(m_sum)))
# x.append(str(float(mood['1']) / float(m_sum)))
# x.append(str(float(mood['2']) / float(m_sum)))
# x.append(str(float(mood['3']) / float(m_sum)))
# x.append(str(float(mood['4']) / float(m_sum)))
# break
file_train.write(y + ' ' + ' '.join(x) + '\n')
def get_prediction_real_data(start, end, out_name):
'''
测评预测结果
'''
# 获取真实大盘数据
results = ts.get_hist_data('sh', start=start, end=end)
print('成功获取 -> 历史大盘数据 ... ...')
# 取得时间序列
series = list(results.index)
print(series)
# 获取数据库连接
coll = mongo_handler.get_stock_coll()
prediction_coll = mongo_handler.get_prediction_coll()
# 遍历每天交易实际数据
for i, date in enumerate(series):
# 第一天留出
if i == 0: continue
# 日期及真实数据
today = datetime.datetime.strptime(date, '%Y-%m-%d').strftime('%Y%m%d')
price_close = results.loc[date]['close']
price_open = results.loc[date]['open']
price_high = results.loc[date]['high']
price_low = results.loc[date]['low']
volume = results.loc[date]['volume'] * 100
last_close = results.loc[series[i - 1]]['close']
extent = results.loc[date]['p_change']
# 涨跌幅当日最大最小
extent_open = (price_open - last_close) / last_close * 100
extent_high = (price_high - last_close) / last_close * 100
extent_low = (price_low - last_close) / last_close * 100
# 创造目标函数
y = list()
y.append(today)
# 收盘价作为预测目标
# y = str(price_close)
# 涨跌幅作为预测目标
# if abs(extent) > 5.0: continue
# y.append(str(extent))
# y.append(str(price_close))
# y.append('1' if extent > 0 else '0')
# y = '1' if extent > 0 else '0'
# 报告中的公式作为预测目标
# y = str(math.log(price_close) - math.log(price_open))
# 其它标准
# y.append(str(extent_high))
# y.append(str(extent_low))
# y.append('1' if extent_open > 0 else '0')
y.append('%.4f' % (extent))
# y.append('%.4f' % (extent_open))
y.append('%.4f' % (extent_high))
y.append('%.4f' % (extent_low))
# y.append(str(volume))
# 获取预测数据
prediction_result = prediction_coll.find_one({'_id': today},
{'_id': 0})
if not prediction_result:
continue
pre_close = prediction_result['model_1']['close']
pre_high = prediction_result['model_1']['high']
pre_low = prediction_result['model_1']['low']
y.append('%.4f' % (pre_close))
y.append('%.4f' % (pre_high))
y.append('%.4f' % (pre_low))
# 收盘方向相同
if (extent * pre_close) > 0:
y.append('1')
else:
y.append('0')
# 数值出现
if pre_close >= extent_low and pre_close <= extent_high:
y.append('1')
else:
y.append('0')
# 最高点方向
if (extent_high * pre_high) > 0:
y.append('1')
else:
y.append('0')
# 最低点方向
if (extent_low * pre_low) > 0:
y.append('1')
else:
y.append('0')
def get_last_mood(dt, n):
'''
内嵌函数
获取前n天的情绪数据
'''
print(dt, n)
dt_n = dec_day(dt, n)
return get_mood_day(coll, dt_n)
x = list()
print('训练数据获取中：', today)
# for i in range(6): # 获取前5天的数据，并且进行预处理
for i in range(1): # 格兰杰, 只取当日
# 获取训练数据日期
train_date = dec_trade_day(today, series, i)
if train_date == False: break
print('训练数据日期：', train_date)
try:
# 一天数据
doc = get_mood_day(coll, train_date)
# 交易时段的数据
# doc = get_mood_15_09(coll, train_date, series)
mood = doc['mood']
# m_sum = doc['stat']['m_sum']
except TypeError:
break
# 绝对情绪
x.append(str(float(mood['0'])))
x.append(str(float(mood['1'])))
x.append(str(float(mood['2'])))
x.append(str(float(mood['3'])))
x.append(str(float(mood['4'])))
# x.append(str(m_sum))
# 绝对情绪修正
# x.append(str(float(mood['0']) / 1000))
# x.append(str(float(mood['1']) / 1000))
# x.append(str(float(mood['2']) / 1000))
# x.append(str(float(mood['3']) / 1000))
# x.append(str(float(mood['4']) / 1000))
# 绝对情绪，取Log
# x.append(str(math.log(float(mood['0']))))
# x.append(str(math.log(float(mood['1']))))
# x.append(str(math.log(float(mood['2']))))
# x.append(str(math.log(float(mood['3']))))
# x.append(str(math.log(float(mood['4']))))
# 数据总和及情绪比例
# x.append(str(float(m_sum) / 100000))
# 总量
five_sum = float(mood['0']) + float(mood['1']) + float(
mood['2']) + float(mood['3']) + float(mood['4'])
x.append(str(float(mood['0']) / five_sum))
x.append(str(float(mood['1']) / five_sum))
x.append(str(float(mood['2']) / five_sum))
x.append(str(float(mood['3']) / five_sum))
x.append(str(float(mood['4']) / five_sum))
# 保留小数点后两位
# x.append('%.4f' % (float(mood['0']) / five_sum))
# x.append('%.4f' % (float(mood['1']) / five_sum))
# x.append('%.4f' % (float(mood['2']) / five_sum))
# x.append('%.4f' % (float(mood['3']) / five_sum))
# x.append('%.4f' % (float(mood['4']) / five_sum))
# x.append(str(float(mood['0']) / float(m_sum)))
# x.append(str(float(mood['1']) / float(m_sum)))
# x.append(str(float(mood['2']) / float(m_sum)))
# x.append(str(float(mood['3']) / float(m_sum)))
# x.append(str(float(mood['4']) / float(m_sum)))
# 各种组合情绪，差值
# x.append( str(float(mood['2']) - float(mood['0'])) )
# x.append( str(float(mood['2']) - float(mood['1'])) )
# x.append( str(float(mood['2']) - float(mood['3'])) )
# x.append( str(float(mood['2']) - float(mood['4'])) )
# 比值
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['0'])))) )
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['1'])))) )
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['2'])))) )
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['3'])))) )
# x.append( str( math.log((1 + float(mood['2'])) / (1 + float(mood['4'])))) )
# 成交量与成交金额，取Log
# x.append(str(math.log(amount)))
# x.append(str(math.log(sum_money)))
else:
# 将训练数据写入到文件
open(out_name, 'a').write(','.join(y) + ',' + ','.join(x) + '\n')