chore: import upstream snapshot with attribution

2026-07-13 13:10:28 +08:00
commit c2acfb5f2d
267 changed files with 15542 additions and 0 deletions
@@ -0,0 +1,4 @@
+## 说明
+
+
+之前测试使用的脚本。执行了一段时间，只是用来进行练习使用的。
@@ -0,0 +1,126 @@
+#!/usr/local/bin/python3
+# -*- coding: utf-8 -*-
+
+
+import libs.common as common
+import pandas as pd
+import numpy as np
+import math
+import datetime
+import heapq
+
+
+### 对每日指标数据，进行筛选。将符合条件的。二次筛选出来。
+def stat_all_lite(tmp_datetime):
+    # 要操作的数据库表名称。
+    table_name = "guess_indicators_lite_buy_daily"
+    datetime_str = (tmp_datetime).strftime("%Y-%m-%d")
+    datetime_int = (tmp_datetime).strftime("%Y%m%d")
+    print("datetime_str:", datetime_str)
+    print("datetime_int:", datetime_int)
+
+    # try:
+    #     # 删除老数据。guess_indicators_lite_buy_daily 是一张单表，没有日期字段。
+    #     del_sql = " DELETE FROM `stock_data`.`%s` WHERE `date`= '%s' " % (table_name, datetime_int)
+    #     print("del_sql:", del_sql)
+    #     common.insert(del_sql)
+    #     print("del_sql")
+    # except Exception as e:
+    #     print("error :", e)
+
+    sql_1 = """
+                SELECT `date`, `code`, `name`, `changepercent`, `trade`,`turnoverratio`, `pb` ,`kdjj`,`rsi_6`,`cci`
+                            FROM stock_data.guess_indicators_lite_daily WHERE `date` = %s 
+                            and `changepercent` > 2 and `pb` > 0 
+        """
+    # and `changepercent` > 2 and `pb` > 0 and `turnoverratio` > 5 去除掉换手率参数。
+    data = pd.read_sql(sql=sql_1, con=common.engine(), params=[datetime_int])
+    data = data.drop_duplicates(subset="code", keep="last")
+    print("######## len data ########:", len(data))
+    # del data["name"]
+    # print(data)
+    data["trade_float32"] = data["trade"].astype('float32', copy=True)
+    # 输入 date 用作历史数据查询。
+    stock_merge = pd.DataFrame({
+        "date": data["date"], "code": data["code"], "wave_mean": data["trade"],
+        "wave_crest": data["trade"], "wave_base": data["trade"]}, index=data.index.values)
+    print(stock_merge.head(1))
+
+    stock_merge = stock_merge.apply(apply_merge, axis=1)  # , axis=1)
+    del stock_merge["date"]  # 合并前删除 date 字段。
+    # 合并数据
+    data_new = pd.merge(data, stock_merge, on=['code'], how='left')
+
+    # 使用 trade_float32 参加计算。
+    data_new = data_new[data_new["trade_float32"] > data_new["wave_base"]]  # 交易价格大于波谷价格。
+    data_new = data_new[data_new["trade_float32"] < data_new["wave_crest"]]  # 小于波峰价格
+
+    # wave_base  wave_crest  wave_mean
+    data_new["wave_base"] = data_new["wave_base"].round(2)  # 数据保留2位小数
+    data_new["wave_crest"] = data_new["wave_crest"].round(2)  # 数据保留2位小数
+    data_new["wave_mean"] = data_new["wave_mean"].round(2)  # 数据保留2位小数
+
+    data_new["up_rate"] = (data_new["wave_mean"].sub(data_new["trade_float32"])).div(data_new["wave_crest"]).mul(100)
+    data_new["up_rate"] = data_new["up_rate"].round(2)  # 数据保留2位小数
+
+    data_new["buy"] = 1
+    data_new["sell"] = 0
+    data_new["today_trade"] = data_new["trade"]
+    data_new["income"] = 0
+    # 重命名 date
+    data_new.columns.values[0] = "buy_date"
+    del data_new["trade_float32"]
+
+    try:
+        common.insert_db(data_new, table_name, False, "`code`")
+        print("insert_db")
+    except Exception as e:
+        print("error :", e)
+    # 重命名
+    del data_new["name"]
+    print(data_new)
+
+
+def apply_merge(tmp):
+    date = tmp["date"]
+    code = tmp["code"]
+    date_end = datetime.datetime.strptime(date, "%Y%m%d")
+    date_start = (date_end + datetime.timedelta(days=-300)).strftime("%Y-%m-%d")
+    date_end = date_end.strftime("%Y-%m-%d")
+    print(code, date_start, date_end)
+
+    # open, high, close, low, volume, price_change, p_change, ma5, ma10, ma20, v_ma5, v_ma10, v_ma20, turnover
+    # 使用缓存方法。加快计算速度。
+    stock = common.get_hist_data_cache(code, date_start, date_end)
+    # 增加空判断，如果是空返回 0 数据。
+    if stock is None:
+        return list([code, date, 0, 0, 0])
+
+    stock = pd.DataFrame({"close": stock["close"]}, index=stock.index.values)
+    stock = stock.sort_index(0)  # 将数据按照日期排序下。
+
+    # print(stock.head(10))
+    arr = pd.Series(stock["close"].values)
+    # print(df_arr)
+    wave_mean = arr.mean()
+    max_point = 3  # 获得最高的几个采样点。
+    # 计算股票的波峰值。
+    wave_crest = heapq.nlargest(max_point, enumerate(arr), key=lambda x: x[1])
+    wave_crest_mean = pd.DataFrame(wave_crest).mean()
+
+    # 输出元祖第一个元素是index，第二元素是比较的数值 计算数据的波谷值
+    wave_base = heapq.nsmallest(max_point, enumerate(arr), key=lambda x: x[1])
+    wave_base_mean = pd.DataFrame(wave_base).mean()
+    # 输出数据
+    print("##############", len(stock))
+    if len(stock) > 180:
+        #     code      date wave_base wave_crest wave_mean 顺序必须一致。返回的是行数据，然后填充。
+        return list([code, date, wave_base_mean[1], wave_crest_mean[1], wave_mean])
+    else:
+        return list([code, date, 0, 0, 0])
+
+
+# main函数入口
+if __name__ == '__main__':
+    # 二次筛选数据。
+    tmp_datetime = common.run_with_args(stat_all_lite)
@@ -0,0 +1,137 @@
+#!/usr/local/bin/python3
+# -*- coding: utf-8 -*-
+
+
+import libs.common as common
+import pandas as pd
+import numpy as np
+import math
+import datetime
+import heapq
+import stockstats
+
+
+# code      date today_trade
+def apply_merge(tmp):
+    date = tmp["date"]
+    code = tmp["code"]
+    date_end = datetime.datetime.strptime(date, "%Y%m%d")
+    date_start = (date_end + datetime.timedelta(days=-300)).strftime("%Y-%m-%d")
+    date_end = date_end.strftime("%Y-%m-%d")
+    print(code, date_start, date_end)
+
+    # open, high, close, low, volume, price_change, p_change, ma5, ma10, ma20, v_ma5, v_ma10, v_ma20, turnover
+    # 使用缓存方法。加快计算速度。
+    stock = common.get_hist_data_cache(code, date_start, date_end)
+    # 增加空判断，如果是空返回 0 数据。
+    if stock is None:
+        return list([code, date, 0.0])
+    print("########")
+    # print(stock.tail(1))
+    close = stock.tail(1)["close"].values[0]
+    print("close:  ", close)
+    print("########")
+    return list([code, date, close])
+
+
+#    buy    code      date  sell  sell_cci  sell_kdjj  sell_rsi_6
+def apply_merge_sell(tmp):
+    date = tmp["date"]
+    code = tmp["code"]
+    date_end = datetime.datetime.strptime(date, "%Y%m%d")
+    date_start = (date_end + datetime.timedelta(days=-300)).strftime("%Y-%m-%d")
+    date_end = date_end.strftime("%Y-%m-%d")
+    print(code, date_start, date_end)
+
+    # open, high, close, low, volume, price_change, p_change, ma5, ma10, ma20, v_ma5, v_ma10, v_ma20, turnover
+    # 使用缓存方法。加快计算速度。
+    stock = common.get_hist_data_cache(code, date_start, date_end)
+    # 增加空判断，如果是空返回 0 数据。
+    if stock is None:
+        return list([1, code, date, 0, 0, 0, 0])
+    print("########")
+    # J大于100时为超买，小于10时为超卖。
+    # 强弱指标保持高于50表示为强势市场，反之低于50表示为弱势市场。
+    # 1、当CCI指标从下向上突破﹢100线而进入非常态区间时，表明股价脱离常态而进入异常波动阶段，
+    # 2、当CCI指标从上向下突破﹣100线而进入另一个非常态区间时，表明股价的盘整阶段已经结束，
+    stockStat = stockstats.StockDataFrame.retype(stock)
+    kdjj = int(stockStat["kdjj"].tail(1).values[0])
+    rsi_6 = int(stockStat["rsi_6"].tail(1).values[0])
+    cci = int(stockStat["cci"].tail(1).values[0])
+    print("kdjj:", kdjj, "rsi_6:", rsi_6, "cci:", cci)
+    # and kdjj > 80 and rsi_6 > 55  and cci > 100 判断卖出时刻。也就是买入时刻的反面。发现有波动就卖了。
+    # if kdjj <= 10 and rsi_6 <= 50 and cci <= 100: old
+    if kdjj <= 80 or rsi_6 <= 55 or cci <= 100:
+        return list([0, code, date, 1, cci, kdjj, rsi_6])
+    else:
+        return list([1, code, date, 0, cci, kdjj, rsi_6])
+
+
+# 增加 收益计算。
+def stat_index_calculate(tmp_datetime):
+    # 要操作的数据库表名称。
+    table_name = "guess_indicators_lite_sell_daily"
+    datetime_str = (tmp_datetime).strftime("%Y-%m-%d")
+    datetime_int = (tmp_datetime).strftime("%Y%m%d")
+    print("datetime_str:", datetime_str)
+    print("datetime_int:", datetime_int)
+
+    sql_1 = """ 
+                SELECT `buy_date`, `code`, `name`, `changepercent`, `trade`, `turnoverratio`, `pb`, `kdjj`, `rsi_6`, 
+                `cci`, `wave_base`, `wave_crest`, `wave_mean`, `up_rate`
+                FROM guess_indicators_lite_buy_daily where `buy_date` <= """ + datetime_int
+    print(sql_1)
+    data = pd.read_sql(sql=sql_1, con=common.engine(), params=[])
+    data = data.drop_duplicates(subset="code", keep="last")
+    print(data["trade"])
+    data["trade_float32"] = data["trade"].astype('float32', copy=False)
+    print(len(data))
+    data["date"] = datetime_int
+
+    stock_merge = pd.DataFrame({
+        "date": data["date"], "code": data["code"], "today_trade": data["trade"]}, index=data.index.values)
+    print(stock_merge.head(1))
+
+    stock_merge = stock_merge.apply(apply_merge, axis=1)  # , axis=1)
+
+    del stock_merge["date"]  # 合并前删除 date 字段。
+    # 合并数据
+    data_new = pd.merge(data, stock_merge, on=['code'], how='left')
+    data_new["income"] = (data_new["today_trade"] - data_new["trade_float32"]) * 100
+    data_new["income"] = data_new["income"].round(4)  # 保留4位小数。
+
+    # 增加售出列。看看是否需要卖出。
+    stock_sell_merge = pd.DataFrame({
+        "date": data["date"], "code": data["code"], "sell": 0, "buy": 0, "sell_kdjj": 0, "sell_rsi_6": 0,
+        "sell_cci": 0},
+        index=data.index.values)
+    print(stock_sell_merge.head(1))
+
+    merge_sell_data = stock_sell_merge.apply(apply_merge_sell, axis=1)  # , axis=1)
+    # 重命名
+    del merge_sell_data["date"]  # 合并前删除 date 字段。
+    # 合并数据
+    data_new = pd.merge(data_new, merge_sell_data, on=['code'], how='left')
+
+    # 删除老数据。
+    try:
+        del_sql = " DELETE FROM `stock_data`.`" + table_name + "` WHERE `date`= '%s' " % datetime_int
+        common.insert(del_sql)
+        print("insert_db")
+    except Exception as e:
+        print("error :", e)
+    del data_new["trade_float32"]
+    try:
+        common.insert_db(data_new, table_name, False, "`date`,`code`")
+        print("insert_db")
+    except Exception as e:
+        print("error :", e)
+    # 重命名
+    del data_new["name"]
+    print(data_new)
+
+
+# main函数入口
+if __name__ == '__main__':
+    # 计算买卖。
+    tmp_datetime = common.run_with_args(stat_index_calculate)
@@ -0,0 +1,119 @@
+#!/usr/local/bin/python3
+# -*- coding: utf-8 -*-
+
+
+import libs.common as common
+import sys
+import time
+import pandas as pd
+import tushare as ts
+from sqlalchemy.types import NVARCHAR
+from sqlalchemy import inspect
+import datetime
+import heapq
+
+"""
+SELECT `date`, `code`, `name`, `changepercent`, `trade`, `open`, `high`, `low`, 
+                `settlement`, `volume`, `turnoverratio`, `amount`, `per`, `pb`, `mktcap`, `nmc` 
+    FROM stock_data.ts_today_all where `date` = 20171106 and trade > 0 and trade <= 20
+and `code` not like '002%' and `code` not like '300%'  and `name` not like '%st%'
+
+"""
+
+
+def stat_index_all(tmp_datetime):
+    datetime_str = (tmp_datetime).strftime("%Y-%m-%d")
+    datetime_int = (tmp_datetime).strftime("%Y%m%d")
+    print("datetime_str:", datetime_str)
+    print("datetime_int:", datetime_int)
+
+    # 查询今日满足股票数据。剔除数据：创业板股票数据，中小板股票数据，所有st股票
+    # #`code` not like '002%' and `code` not like '300%'  and `name` not like '%st%'
+    sql_1 = """ 
+            SELECT `date`, `code`, `name`, `changepercent`, `trade`, `open`, `high`, `low`, 
+                `settlement`, `volume`, `turnoverratio`, `amount`, `per`, `pb`, `mktcap`, `nmc` 
+            FROM stock_data.ts_today_all WHERE `date` = %s and `trade` > 0 and `open` > 0 and trade <= 20 
+                and `code` not like %s and `code` not like %s and `name` not like %s
+            """
+    print(sql_1)
+    data = pd.read_sql(sql=sql_1, con=common.engine(), params=[datetime_int, '002%', '300%', '%st%'])
+    print(type(data))
+    data = data.drop_duplicates(subset="code", keep="last")
+    print(data["trade"])
+    data["trade_float32"] = data["trade"].astype('float32', copy=False)
+    print(len(data))
+    print("########data[trade]########:")
+    print(data["trade"])
+
+    # 使用 trade 填充数据
+    stock_guess = pd.DataFrame({
+        "date": data["date"], "code": data["code"], "wave_mean": data["trade"],
+        "wave_crest": data["trade"], "wave_base": data["trade"]}, index=data.index.values)
+    print(stock_guess.head())
+    stock_guess = stock_guess.apply(apply_guess, axis=1)  # , axis=1)
+    print(stock_guess.head())
+    # stock_guess.astype('float32', copy=False)
+    stock_guess.drop('date', axis=1, inplace=True)  # 删除日期字段，然后和原始数据合并。
+    stock_guess = stock_guess.round(2)  # 数据保留2位小数
+    print(stock_guess["wave_base"])
+
+    data_new = pd.merge(data, stock_guess, on=['code'], how='left')
+    print("#############")
+
+    # 使用pandas 函数 ： https://pandas.pydata.org/pandas-docs/stable/api.html#id4
+    data_new["up_rate"] = (data_new["trade_float32"].sub(data_new["wave_mean"])).div(data_new["wave_crest"]).mul(100)
+    data_new["up_rate"] = data_new["up_rate"].round(2)  # 数据保留2位小数
+    data_new.drop('trade_float32', axis=1, inplace=True)  # 删除计算字段。
+
+    # 删除老数据。
+    del_sql = " DELETE FROM `stock_data`.`guess_period_daily` WHERE `date`= '%s' " % datetime_int
+    common.insert(del_sql)
+    # print(data_new.head())
+    # data_new["down_rate"] = (data_new["trade"] - data_new["wave_mean"]) / data_new["wave_base"]
+    common.insert_db(data_new, "guess_period_daily", False, "`date`,`code`")
+
+    # 进行左连接.
+    # tmp = pd.merge(tmp, tmp2, on=['company_id'], how='left')
+
+
+def apply_guess(tmp):
+    date = tmp["date"]
+    code = tmp["code"]
+    date_end = datetime.datetime.strptime(date, "%Y%m%d")
+    date_start = (date_end + datetime.timedelta(days=-300)).strftime("%Y-%m-%d")
+    date_end = date_end.strftime("%Y-%m-%d")
+    print(code, date_start, date_end)
+
+    # open, high, close, low, volume, price_change, p_change, ma5, ma10, ma20, v_ma5, v_ma10, v_ma20, turnover
+    # 使用缓存方法。加快计算速度。
+    stock = common.get_hist_data_cache(code, date_start, date_end)
+    # 增加空判断，如果是空返回 0 数据。
+    if stock is None:
+        return pd.Series([date, code, 0.0, 0.0, 0.0],
+                     index=['date', 'code', 'wave_mean', 'wave_crest', 'wave_base'])
+
+    stock = pd.DataFrame({"close": stock["close"]}, index=stock.index.values)
+    stock = stock.sort_index(0)  # 将数据按照日期排序下。
+
+    # print(stock.head(10))
+    arr = pd.Series(stock["close"].values)
+    # print(df_arr)
+    wave_mean = arr.mean()
+    # 计算股票的波峰值。
+    wave_crest = heapq.nlargest(5, enumerate(arr), key=lambda x: x[1])
+    wave_crest_mean = pd.DataFrame(wave_crest).mean()
+
+    # 输出元祖第一个元素是index，第二元素是比较的数值 计算数据的波谷值
+    wave_base = heapq.nsmallest(5, enumerate(arr), key=lambda x: x[1])
+    wave_base_mean = pd.DataFrame(wave_base).mean()
+    # 输出数据
+    # print("##############")
+    #     code      date wave_base wave_crest wave_mean 顺序必须一致。返回的是行数据，然后填充。
+    return pd.Series([date, code, wave_base_mean[1], wave_crest_mean[1], wave_mean],
+                     index=['date','code','wave_mean','wave_crest','wave_base'])
+
+
+# main函数入口
+if __name__ == '__main__':
+    # 使用方法传递。
+    tmp_datetime = common.run_with_args(stat_index_all)
@@ -0,0 +1,130 @@
+#!/usr/local/bin/python3
+# -*- coding: utf-8 -*-
+
+
+import libs.common as common
+import sys
+import time
+import pandas as pd
+import numpy as np
+import math
+import tushare as ts
+from sqlalchemy.types import NVARCHAR
+from sqlalchemy import inspect
+import datetime
+import heapq
+
+"""
+SELECT `date`, `code`, `name`, `changepercent`, `trade`, `open`, `high`, `low`, 
+                `settlement`, `volume`, `turnoverratio`, `amount`, `per`, `pb`, `mktcap`, `nmc` 
+    FROM stock_data.ts_today_all where `date` = 20171106 and trade > 0 and trade <= 20
+and `code` not like '002%' and `code` not like '300%'  and `name` not like '%st%'
+
+"""
+
+
+def stat_index_all(tmp_datetime):
+    datetime_str = (tmp_datetime).strftime("%Y-%m-%d")
+    datetime_int = (tmp_datetime).strftime("%Y%m%d")
+    print("datetime_str:", datetime_str)
+    print("datetime_int:", datetime_int)
+
+    # 查询今日满足股票数据。剔除数据：创业板股票数据，中小板股票数据，所有st股票
+    # #`code` not like '002%' and `code` not like '300%'  and `name` not like '%st%'
+    sql_1 = """ 
+            SELECT `date`, `code`, `name`, `changepercent`, `trade`, `open`, `high`, `low`, 
+                `settlement`, `volume`, `turnoverratio`, `amount`, `per`, `pb`, `mktcap`, `nmc` 
+            FROM stock_data.ts_today_all WHERE `date` = %s and `trade` > 0 and `open` > 0 and trade <= 20 
+                and `code` not like %s and `code` not like %s and `name` not like %s
+            """
+    print(sql_1)
+    data = pd.read_sql(sql=sql_1, con=common.engine(), params=[datetime_int, '002%', '300%', '%st%'])
+    data = data.drop_duplicates(subset="code", keep="last")
+    print("########data[trade]########:")
+    # print(data["trade"])
+
+    # 使用 trade 填充数据
+    stock_guess = pd.DataFrame({
+        "date": data["date"], "code": data["code"], "5d": data["trade"],
+        "10d": data["trade"], "20d": data["trade"], "60d": data["trade"], "5-10d": data["trade"],
+        "5-20d": data["trade"], "return": data["trade"], "mov_vol": data["trade"]
+    }, index=data.index.values)
+
+    stock_guess = stock_guess.apply(apply_guess, axis=1)  # , axis=1)
+    # print(stock_guess.head())
+    # stock_guess.astype('float32', copy=False)
+    stock_guess.drop('date', axis=1, inplace=True)  # 删除日期字段，然后和原始数据合并。
+
+    # print(stock_guess["5d"])
+
+    data_new = pd.merge(data, stock_guess, on=['code'], how='left')
+    print("#############")
+
+    # 使用pandas 函数 ： https://pandas.pydata.org/pandas-docs/stable/api.html#id4
+    data_new["return"] = data_new["return"].mul(100)  # 扩大100 倍方便观察
+    data_new["mov_vol"] = data_new["mov_vol"].mul(100)
+
+    data_new = data_new.round(2)  # 数据保留2位小数
+
+    # 删除老数据。
+    del_sql = " DELETE FROM `stock_data`.`guess_return_daily` WHERE `date`= '%s' " % datetime_int
+    common.insert(del_sql)
+
+    # data_new["down_rate"] = (data_new["trade"] - data_new["wave_mean"]) / data_new["wave_base"]
+    common.insert_db(data_new, "guess_return_daily", False, "`date`,`code`")
+
+    # 进行左连接.
+    # tmp = pd.merge(tmp, tmp2, on=['company_id'], how='left')
+
+
+def apply_guess(tmp):
+    date = tmp["date"]
+    code = tmp["code"]
+    date_end = datetime.datetime.strptime(date, "%Y%m%d")
+    date_start = (date_end + datetime.timedelta(days=-300)).strftime("%Y-%m-%d")
+    date_end = date_end.strftime("%Y-%m-%d")
+    print(code, date_start, date_end)
+    # open, high, close, low, volume, price_change, p_change, ma5, ma10, ma20, v_ma5, v_ma10, v_ma20, turnover
+    # 使用缓存方法。加快计算速度。
+    stock = common.get_hist_data_cache(code, date_start, date_end)
+    # 增加空判断，如果是空返回 0 数据。
+    if stock is None:
+        return pd.Series([0.0, 0.0, 0.0, 0.0, 0.0, 0.0, code, date, 0.0, 0.0],
+                         index=['10d', '20d', '5-10d', '5-20d', '5d', '60d', 'code', 'date', 'mov_vol', 'return'])
+
+    stock = pd.DataFrame({"close": stock["close"]}, index=stock.index.values)
+    stock = stock.sort_index(0)  # 将数据按照日期排序下。
+    # print(stock.head(10))
+    # 5周期、10周期、20周期和60周期
+    # 周线、半月线、月线和季度线
+    stock["5d"] = stock["close"].rolling(window=5).mean()  # 周线
+    stock["10d"] = stock["close"].rolling(window=10).mean()  # 半月线
+    stock["20d"] = stock["close"].rolling(window=20).mean()  # 月线
+    stock["60d"] = stock["close"].rolling(window=60).mean()  # 季度线
+    # 计算日期差。
+    stock["5-10d"] = (stock["5d"] - stock["10d"]) * 100 / stock["10d"]  # 周-半月线差
+    stock["5-20d"] = (stock["5d"] - stock["20d"]) * 100 / stock["20d"]  # 周-月线差
+    # 计算股票的收益价格
+    stock["return"] = np.log(stock["close"] / stock["close"].shift(1))
+
+    # print(stock["return"])
+    # 计算股票的【收益率的移动历史标准差】
+    mov_day = int(len(stock) / 20)
+    # print("mov_day:", mov_day, len(stock))
+    stock["mov_vol"] = stock["return"].rolling(window=mov_day).std() * math.sqrt(mov_day)
+    # print(stock["mov_vol"].tail())
+    # print(stock["return"].tail())
+    # print("stock[10d].tail(1)", stock["10d"].tail(1).values[0])
+    # 10d    20d  5-10d  5-20d     5d    60d    code      date  mov_vol  return
+    tmp = pd.Series([stock["10d"].tail(1).values[0], stock["20d"].tail(1).values[0], stock["5-10d"].tail(1).values[0],
+                stock["5-20d"].tail(1).values[0], stock["5d"].tail(1).values[0], stock["60d"].tail(1).values[0],
+                code, date, stock["mov_vol"].tail(1).values[0], stock["return"].tail(1).values[0]],
+                    index=['10d', '20d', '5-10d', '5-20d', '5d', '60d', 'code', 'date', 'mov_vol', 'return'])
+    # print(tmp)
+    return tmp
+
+
+# main函数入口
+if __name__ == '__main__':
+    # 使用方法传递。
+    tmp_datetime = common.run_with_args(stat_index_all)
@@ -0,0 +1,146 @@
+#!/usr/local/bin/python3
+# -*- coding: utf-8 -*-
+
+
+import libs.common as common
+import pandas as pd
+import numpy as np
+import math
+import datetime
+import sklearn as skl
+from sklearn import datasets, linear_model
+# https://github.com/udacity/machine-learning/issues/202
+# sklearn.cross_validation 这个包不推荐使用了。
+from sklearn.model_selection import train_test_split, cross_val_score
+from sklearn.neighbors import KNeighborsClassifier
+
+# 要操作的数据库表名称。
+table_name = "guess_sklearn_ma_daily"
+
+
+# 批处理数据。
+def stat_all_batch(tmp_datetime):
+    datetime_str = (tmp_datetime).strftime("%Y-%m-%d")
+    datetime_int = (tmp_datetime).strftime("%Y%m%d")
+    print("datetime_str:", datetime_str)
+    print("datetime_int:", datetime_int)
+
+    try:
+        # 删除老数据。
+        del_sql = " DELETE FROM `stock_data`.`%s` WHERE `date`= %s " % (table_name, datetime_int)
+        print("del_sql:", del_sql)
+        common.insert(del_sql)
+    except Exception as e:
+        print("error :", e)
+
+    sql_count = """
+            SELECT count(1) FROM stock_data.ts_today_all WHERE `date` = %s and `trade` > 0 and `open` > 0 and trade <= 20 
+                 and `code` not like %s and `name` not like %s 
+    """
+    # 修改逻辑，增加中小板块计算。 中小板：002，创业板：300 。and `code` not like %s and `code` not like %s and `name` not like %s
+    # count = common.select_count(sql_count, params=[datetime_int, '002%', '300%', '%st%'])
+    count = common.select_count(sql_count, params=[datetime_int, '300%', '%st%'])
+    print("count :", count)
+    batch_size = 100
+    end = int(math.ceil(float(count) / batch_size) * batch_size)
+    print(end)
+    # for i in range(0, end, batch_size):
+    for i in range(0, end, batch_size):
+        print("loop :", i)
+        # 查询今日满足股票数据。剔除数据：创业板股票数据，中小板股票数据，所有st股票
+        # #`code` not like '002%' and `code` not like '300%'  and `name` not like '%st%'
+        sql_1 = """ 
+                    SELECT `date`, `code`, `name`, `changepercent`, `trade`, `open`, `high`, `low`, 
+                        `settlement`, `volume`, `turnoverratio`, `amount`, `per`, `pb`, `mktcap`, `nmc` 
+                    FROM stock_data.ts_today_all WHERE `date` = %s and `trade` > 0 and `open` > 0 and trade <= 20 
+                        and `code` not like %s and `name` not like %s limit %s , %s
+                    """
+        print(sql_1)
+        # data = pd.read_sql(sql=sql_1, con=common.engine(), params=[datetime_int, '002%', '300%', '%st%', i, batch_size])
+        data = pd.read_sql(sql=sql_1, con=common.engine(), params=[datetime_int, '300%', '%st%', i, batch_size])
+        data = data.drop_duplicates(subset="code", keep="last")
+        print("########data[trade]########:", len(data))
+
+        # 使用 trade 填充数据
+        stock_sklearn = pd.DataFrame({
+            "date": data["date"], "code": data["code"], "next_close": data["trade"],
+            "sklearn_score": data["trade"]}, index=data.index.values)
+        print(stock_sklearn.head())
+        stock_sklearn_apply = stock_sklearn.apply(apply_sklearn, axis=1)  # , axis=1)
+        # 重命名
+        del stock_sklearn_apply["date"]  # 合并前删除 date 字段。
+        # 合并数据
+        data_new = pd.merge(data, stock_sklearn_apply, on=['code'], how='left')
+        # for index, row in data.iterrows():
+        #     next_stock, score = stat_index_all(row, i)
+        #     print(next_stock, score)
+        data_new["next_close"] = data_new["next_close"].round(2)  # 数据保留4位小数
+        data_new["sklearn_score"] = data_new["sklearn_score"].round(2)  # 数据保留2位小数
+
+        data_new["trade_float32"] = data["trade"].astype('float32', copy=False)
+        data_new["up_rate"] = (data_new["next_close"] - data_new["trade_float32"]) * 100 / data_new["trade_float32"]
+        data_new["up_rate"] = data_new["up_rate"].round(2)  # 数据保留2位小数
+        del data_new["trade_float32"]
+
+        try:
+            common.insert_db(data_new, table_name, False, "`date`,`code`")
+            print("insert_db")
+        except Exception as e:
+            print("error :", e)
+        # 重命名
+        del data_new["name"]
+        print(data_new)
+
+
+# code date next_close sklearn_score
+def apply_sklearn(data):
+    # 要操作的数据库表名称。
+    print("########stat_index_all########:", len(data))
+    date = data["date"]
+    code = data["code"]
+    print(date, code)
+    date_end = datetime.datetime.strptime(date, "%Y%m%d")
+    date_start = (date_end + datetime.timedelta(days=-300)).strftime("%Y-%m-%d")
+    date_end = date_end.strftime("%Y-%m-%d")
+    print(code, date_start, date_end)
+
+    # open high close low volume price_change p_change ma5 ma10 ma20 v_ma5 v_ma10 v_ma20 turnover
+    stock_X = common.get_hist_data_cache(code, date_start, date_end)
+    # 增加空判断，如果是空返回 0 数据。
+    if stock_X is None:
+        return list([code, date, 0.0, 0.0])
+
+    stock_X = stock_X.sort_index(0)  # 将数据按照日期排序下。
+    stock_y = pd.Series(stock_X["close"].values)  # 标签
+
+    stock_X_next = stock_X.iloc[len(stock_X) - 1]
+    print("########################### stock_X_next date:", stock_X_next)
+    # 使用今天的交易价格，13 个指标预测明天的价格。偏移股票数据，今天的数据，目标是明天的价格。
+    stock_X = stock_X.drop(stock_X.index[len(stock_X) - 1])  # 删除最后一条数据
+    stock_y = stock_y.drop(stock_y.index[0])  # 删除第一条数据
+    # print("########################### stock_X date:", stock_X)
+
+    # 删除掉close 也就是收盘价格。
+    del stock_X["close"]
+    del stock_X_next["close"]
+
+    model = linear_model.LinearRegression()
+    # model = KNeighborsClassifier()
+
+    model.fit(stock_X.values, stock_y)
+    # print("############## test_akshare & target #############")
+    # print("############## coef_ & intercept_ #############")
+    # print(model.coef_)  # 系数
+    # print(model.intercept_)  # 截断
+    next_close = model.predict([stock_X_next.values])
+    if len(next_close) == 1:
+        next_close = next_close[0]
+    sklearn_score = model.score(stock_X.values, stock_y)
+    print("score:", sklearn_score)  # 评分
+    return list([code, date, next_close, sklearn_score * 100])
+
+
+# main函数入口
+if __name__ == '__main__':
+    # 使用方法传递。
+    tmp_datetime = common.run_with_args(stat_all_batch)