GitHub - jm199504/Financial-Knowledge-Graphs: 小型金融知识图谱构建流程

import tushare as ts
import csv
import time
import pandas as pd
# 以下pro_api token可能已过期，可自行前往申请或者使用免费版本
pro = ts.pro_api('4340a981b3102106757287c11833fc14e310c4bacf8275f067c9b82d')

stock_basic = pro.stock_basic(list_status='L', fields='ts_code, symbol, name, industry')
# 重命名行，便于后面导入neo4j
basic_rename = {'ts_code': 'TS代码', 'symbol': '股票代码', 'name': '股票名称', 'industry': '行业'}
stock_basic.rename(columns=basic_rename, inplace=True)
# 保存为stock_basic.csv
stock_basic.to_csv('financial_data\\stock_basic.csv', encoding='gbk')

holders = pd.DataFrame(columns=('ts_code', 'ann_date', 'end_date', 'holder_name', 'hold_amount', 'hold_ratio'))
# 获取一年内所有上市股票股东信息（可以获取一个报告期的）
for i in range(3610):
   code = stock_basic['TS代码'].values[i]
   holders = pro.top10_holders(ts_code=code, start_date='20180101', end_date='20181231')
   holders = holders.append(holders)
   if i % 600 == 0:
       print(i)
   time.sleep(0.4)# 数据接口限制
# 保存为stock_holders.csv
holders.to_csv('financial_data\\stock_holders.csv', encoding='gbk')
holders = pro.holders(ts_code='000001.SZ', start_date='20180101', end_date='20181231')

concept_details = pd.DataFrame(columns=('id', 'concept_name', 'ts_code', 'name'))
for i in range(358):
   id = 'TS' + str(i)
   concept_detail = pro.concept_detail(id=id)
   concept_details = concept_details.append(concept_detail)
   time.sleep(0.4)
# 保存为concept_detail.csv
concept_details.to_csv('financial_data\\stock_concept.csv', encoding='gbk')

for i in range(3610):
   code = stock_basic['TS代码'].values[i]
   notices = pro.anns(ts_code=code, start_date='20180101', end_date='20181231', year='2018')
   notices.to_csv("financial_data\\notices\\"+str(code)+".csv",encoding='utf_8_sig',index=False)
notices = pro.anns(ts_code='000001.SZ', start_date='20180101', end_date='20181231', year='2018')

news = pro.news(src='sina', start_date='20180101', end_date='20181231')
news.to_csv("financial_data\\news.csv",encoding='utf_8_sig')

concept = pro.concept()
concept.to_csv('financial_data\\concept.csv', encoding='gbk')

#获取沪股通成分
sh = pro.hs_const(hs_type='SH')
sh.to_csv("financial_data\\sh.csv",index=False)
#获取深股通成分
sz = pro.hs_const(hs_type='SZ')
sz.to_csv("financial_data\\sz.csv",index=False)

for i in range(3610):
   code = stock_basic['TS代码'].values[i]
   price = pro.query('daily', ts_code=code, start_date='20180101', end_date='20181231')
   price.to_csv("financial_data\\price\\"+str(code)+".csv",index=False)

# 基本面信息
df = ts.get_stock_basics()
# 公告信息
ts.get_notices("000001")
# 新浪股吧
ts.guba_sina()
# 历史价格数据
ts.get_hist_data("000001")
# 历史价格数据（周粒度）
ts.get_hist_data("000001",ktype="w")
# 历史价格数据（1分钟粒度）
ts.get_hist_data("000001",ktype="m")
# 历史价格数据（5分钟粒度）
ts.get_hist_data("000001",ktype="5")
# 指数数据（sh上证指数;sz深圳成指;hs300沪深300;sz50上证50;zxb中小板指数;cyb创业板指数）
ts.get_hist_data("cyb")
# 宏观数据(居民消费指数)
ts.get_cpi()
# 获取分笔数据
ts.get_tick_data('000001', date='2018-10-08', src='tt')

import numpy as np

yaxis = list()
for i in listdir:
    stock = pd.read_csv("financial_data\\price_logreturn\\"+i)
    yaxis.append(len(stock['logreturn']))
counts = np.bincount(yaxis)

np.argmax(counts)

import pandas as pd
import numpy as np
import os
import math

listdir = os.listdir("financial_data\\price")

for l in listdir:
   stock = pd.read_csv('financial_data\\price\\'+l)
   stock['index'] = [1]* len(stock['close'])
   stock['next_close'] = stock.groupby('index')['close'].shift(-1)
   stock = stock.drop(index=stock.index[-1])
   logreturn = list()
   for i in stock.index:
       logreturn.append(math.log(stock['next_close'][i]/stock['close'][i]))
   stock['logreturn'] = logreturn
   stock.to_csv("financial_data\\price_logreturn\\"+l,index=False)

from math import sqrt
def multipl(a,b):
   sumofab=0.0
   for i in range(len(a)):
       temp=a[i]*b[i]
       sumofab+=temp
   return sumofab

def corrcoef(x,y):
   n=len(x)
   #求和
   sum1=sum(x)
   sum2=sum(y)
   #求乘积之和
   sumofxy=multipl(x,y)
   #求平方和
   sumofx2 = sum([pow(i,2) for i in x])
   sumofy2 = sum([pow(j,2) for j in y])
   num=sumofxy-(float(sum1)*float(sum2)/n)
   #计算皮尔逊相关系数
   den=sqrt((sumofx2-float(sum1**2)/n)*(sumofy2-float(sum2**2)/n))
   return num/den

listdir = os.listdir("financial_data\\300stock_logreturn")
s1 = list()
s2 = list()
corr = list()
for i in listdir:
   for j in listdir:
       stocka = pd.read_csv("financial_data\\300stock_logreturn\\"+i)
       stockb = pd.read_csv("financial_data\\300stock_logreturn\\"+j)
       if len(stocka['logreturn']) == 242 and len(stockb['logreturn']) == 242:
           s1.append(str(i)[:10])
           s2.append(str(j)[:10])
           corr.append(corrcoef(stocka['logreturn'],stockb['logreturn']))
           print(str(i)[:10],str(j)[:10],corrcoef(stocka['logreturn'],stockb['logreturn']))
corrdf = pd.DataFrame()
corrdf['s1'] = s1
corrdf['s2'] = s2
corrdf['corr'] = corr
corrdf.to_csv("financial_data\\corr.csv")

from pandas import DataFrame
from py2neo import Graph,Node,Relationship,NodeMatcher
import pandas as pd
import numpy as np
import os
# 连接Neo4j数据库
graph = Graph('http://localhost:7474/db/data/',username='neo4j',password='neo4j')

stock = pd.read_csv('stock_basic.csv',encoding="gbk")
holder = pd.read_csv('holders.csv')
concept_num = pd.read_csv('concept.csv')
concept = pd.read_csv('stock_concept.csv')
sh = pd.read_csv('sh.csv')
sz = pd.read_csv('sz.csv')
corr = pd.read_csv('corr.csv')

stock['行业'] = stock['行业'].fillna('未知')
holder = holder.drop_duplicates(subset=None, keep='first', inplace=False)

sz = Node('深股通',名字='深股通')
graph.create(sz)  

sh = Node('沪股通',名字='沪股通')
graph.create(sh)  

for i in concept_num.values:
   a = Node('概念',概念代码=i[1],概念名称=i[2])
   print('概念代码:'+str(i[1]),'概念名称:'+str(i[2]))
   graph.create(a)

for i in stock.values:
   a = Node('股票',TS代码=i[1],股票名称=i[3],行业=i[4])
   print('TS代码:'+str(i[1]),'股票名称:'+str(i[3]),'行业:'+str(i[4]))
   graph.create(a)

for i in holder.values:
   a = Node('股东',TS代码=i[0],股东名称=i[1],持股数量=i[2],持股比例=i[3])
   print('TS代码:'+str(i[0]),'股东名称:'+str(i[1]),'持股数量:'+str(i[2]))
   graph.create(a)

matcher = NodeMatcher(graph)
for i in holder.values:    
   a = matcher.match("股票",TS代码=i[0]).first()
   b = matcher.match("股东",TS代码=i[0])
   for j in b:
       r = Relationship(j,'参股',a)
       graph.create(r)
       print('TS',str(i[0]))
           
for i in concept.values:
   a = matcher.match("股票",TS代码=i[3]).first()
   b = matcher.match("概念",概念代码=i[1]).first()
   if a == None or b == None:
       continue
   r = Relationship(a,'概念属于',b)
   graph.create(r)

noticesdir = os.listdir("notices\\")
for n in noticesdir:
   notice = pd.read_csv("notices\\"+n,encoding="utf_8_sig")
   notice['content'] = notice['content'].fillna('空白')
   for i in notice.values:
       a = matcher.match("股票",TS代码=i[0]).first()
       b = Node('公告',日期=i[1],标题=i[2],内容=i[3])
       graph.create(b)
       r = Relationship(a,'发布公告',b)
       graph.create(r)
       print(str(i[0]))
       
for i in sz.values:
   a = matcher.match("股票",TS代码=i[0]).first()
   b = matcher.match("深股通").first()
   r = Relationship(a,'成分股属于',b)
   graph.create(r)
   print('TS代码:'+str(i[1]),'--深股通')

for i in sh.values:
   a = matcher.match("股票",TS代码=i[0]).first()
   b = matcher.match("沪股通").first()
   r = Relationship(a,'成分股属于',b)
   graph.create(r)
   print('TS代码:'+str(i[1]),'--沪股通')

# 构建股票间关联
corr = pd.read_csv("corr.csv")
for i in corr.values:
   a = matcher.match("股票",TS代码=i[1][:-1]).first()
   b = matcher.match("股票",TS代码=i[2][:-1]).first()
   r = Relationship(a,str(i[3]),b)
   graph.create(r)
   print(i)