【量化投资】策略六（聚宽）

简述

相比之前的策略4，这里就是换成了用一个新的因子。两者对比来看，似乎发现是不是我的beta对冲的时候出现了问题。

由于这里我大多数数据都换成用scipy和numpy来算，所以计算速度明显增加了。

对比着来看，这个因子似乎是比之前的策略4用的alpha1要好很多。虽然最终的结果只差一个小数点。但是我觉得很有可能是出现在那个操作股指期货的方式上了。
这个可以有待改进。

效果图

代码

# 导入函数库
import statsmodels.api as sm
from statsmodels import regression
import numpy as np
import pandas as pd
import time
from datetime import date
from jqdata import *
from jqlib.alpha191 import *
import scipy.stats as stats


# 一键回测说明：
# 百度聚宽-》注册账号-》我的策略里面创建策略-》复制代码到里面
# 右边回测 开始时间：2017-1-1 终止时间：今天 资金：10000000

# 初始化函数，设定基准等等
def initialize(context):
    g.tc = 2  # 调仓频率

    # 下面是框架固定部分，不需要修改    
    g.N = 50  # 持仓数目

    g.t = 0  # 记录运行的天数
    g.weight_list = [1]  # 因子的权重参数
    log.set_level('order', 'error')
    set_option('use_real_price', True)  # 用真实价格交易
    set_slippage(FixedSlippage(0))  # 将滑点设置为0
    set_commission(PerTrade(buy_cost=0.0000, sell_cost=0.000, min_cost=0))  # 手续费设置为0
    # set_benchmark('000905.XSHG')  #中证500为业绩基准
    set_benchmark('000300.XSHG')  # 沪深300为业绩基准
    # 选股范围为全市场选股：上证+深证股票
    # g.stockrange= get_index_stocks('000001.XSHG')+get_index_stocks('399106.XSHE')
    g.stockrange = get_index_stocks('000300.XSHG')

    init_cash = context.portfolio.starting_cash/2

    # 操作期货
    set_subportfolios([SubPortfolioConfig(cash=init_cash, type='stock'),
                       SubPortfolioConfig(cash=init_cash, type='futures')])

    # 多少比例的资金用来对冲beta
    g.rate = 0.5
    # 历史最高收益率
    g.returnsRate = 0
    # 调整临界限比率
    g.changeLimiteRate = 0.2
    # 调整比例
    g.changeRate = 0


    # 
    g.betaName = 'IF9999.CCFX'

## 每根日线运行一次
def handle_data(context, data):
    if g.t == 0:
        # 设置可行股票池：用set_feasible_stocks函数剔除当前或者计算样本期间停牌的股票
        g.all_stocks = pickStock(context, g.stockrange)
        previousStr = str(context.previous_date)

        # adjustedBeta
        adjustedBeta(context)

        stock_sort = get_all_cleaned_factor_ranked(g.all_stocks, g.weight_list, previousStr)
        # 调仓
        rebalance_position(context, stock_sort)

        # order_stock_sell(context,data,stock_sort)
        # order_stock_buy(context,data,stock_sort)       

    # g.t+=1 # speed up
    g.t = (g.t + 1) % g.tc


def get_all_cleaned_factor_ranked(stocks, weight_list, previousStr):
    # value = my_apha1(stocks)
    value = my_alpha2(stocks)
    value = (-value).argsort()[:g.N]
    return list(map(lambda x: stocks[x], value))

'''
调整对冲beta的量
'''

def adjustedBeta(context):
    current_returns = context.portfolio.returns

    if current_returns <= g.returnsRate * ( 1 - g.changeLimiteRate):
        g.changeRate += 1
        g.rate = 0.5 + (0.5 / (1+np.exp(-g.changeRate)))
    elif current_returns >= g.returnsRate * ( 1 + g.changeLimiteRate):
        g.changeRate -= 1
        g.rate = 0.5 + (0.5 / (1+np.exp(-g.changeRate)))

    if current_returns > g.returnsRate:
        g.returnsRate = current_returns

'''
END 调整对冲beta的量
'''




'''
factor1
'''


def single_alpha1(stock):
    df = get_price(stock, count=26, fields=['close'])
    returns = (df[1:] - df[:-2]) / df[:-2]  # 25 lines
    value = [0] * 5
    for i in range(5):
        if np.array(returns).tolist()[-1 - i] < 0:
            value[i] = returns.ix[-20 - i:-1 - i].std() ** 2
        else:
            value[i] = df['close'][-1 - i] ** 2
    return 5 - np.array(value).argmax()


def my_apha1(stocks):
    value = list(map(single_alpha1, stocks))
    MAX = max(value)
    MIN = min(value)

    value = (np.array(value) - MIN) / (MAX - MIN)
    return value


'''
END factor1
'''


'''
factor2
'''

def single_alpha2(stock):
    df = get_price(stock, count=8, fields=['open', 'close', 'volume'])
    volume = np.log(df['volume'].tolist())
    delta_volume = volume[2:] - volume[:-2]
    close = df['close']
    open_ = df['open']
    rate = (close - open_) / open_
    rate = np.array(rate)[2:]
    return delta_volume, rate

def my_alpha2(stocks):
    values = list(map(single_alpha2, stocks))
    T_values = list(zip(*values))
    T_values = list(map(doubleListRank, T_values))
    T_values = list(map(list, zip(*T_values)))
    values = np.array(map(Neg_correlation, T_values))
    return values

def Neg_correlation(adlist):
    return -stats.pearsonr(np.array(adlist[0]), np.array(adlist[1]))[0]

def doubleListRank(ddlist):
    ddlist6 = list(zip(*ddlist))
    ddlist6 = list(map(rank, ddlist6))
    return list(map(list, zip(*ddlist6)))

def rank(alist):
    MAX = max(alist)
    MIN = min(alist)
    return (np.array(alist) - MIN) / (MAX - MIN)


# (-1 * correlation(rank(delta(log(volume), 2)), rank(((close - open)/ open)), 6))
# 6天以来的相关系数() 两个数值分别是
# rank(delta(log(volume), 2))

'''
END factor2
'''


'''
初步筛选股票
'''


# pick stocks
def pickStock(context, stocks):
    # universe = set_feasible_stocks(stocks)
    universe = filter_specials(stocks, context)
    # 过滤上市时间小于60天的股票
    for stock in universe:
        days_public = (context.current_dt.date() - get_security_info(stock).start_date).days
        if days_public < 60:
            universe.remove(stock)
    g.lenth = len(universe)
    return universe


# 过滤停牌的股票
def set_feasible_stocks(stock_list):
    current_data = get_current_data()
    stock_list = [stock for stock in stock_list if not current_data[stock].paused]  # 不考虑停盘的股票
    return stock_list


# 将多因子的dataframe进行排序，并且将有空值的行去掉
def rank_stock(all_factor, weight_list):
    C = len(all_factor.columns)
    ranked = all_factor.iloc[:, 0].rank() * weight_list[0]
    if C > 1:
        for j in range(1, C):
            ranked = all_factor.iloc[:, j].rank() * weight_list[j] + ranked
    ranked = pd.DataFrame(ranked)
    ranked.columns = ['rank']
    one_sort = ranked.sort('rank', ascending=g.ascending)
    stock_sort = one_sort.index[:g.N]
    return stock_sort


#过滤退市，停牌，ST

def filter_specials(stock_list,context):
    curr_data = get_current_data()
    stock_list = [stock for stock in stock_list if \
                  (not curr_data[stock].paused)  # 未停牌
                  and (not curr_data[stock].is_st)  # 非ST
                  and ('ST' not in curr_data[stock].name)
                  and ('*' not in curr_data[stock].name)
                  and ('退' not in curr_data[stock].name)
                  and (curr_data[stock].low_limit < curr_data[stock].day_open < curr_data[stock].high_limit)                   ]
    return stock_list


'''
END 初步筛选股票
'''

'''
调仓
'''


def rebalance_position(context, stocks_list):
    current_holding = context.subportfolios[0].positions.keys()
    stocks_to_sell = list(set(current_holding) - set(stocks_list))
    # 卖出
    bulk_orders(stocks_to_sell, 0)
    total_value = context.subportfolios[0].total_value
    set_option('futures_margin_rate', g.rate)
    bete_order(context, (1 + g.rate) * context.subportfolios[1].total_value)
    # 买入  
    bulk_orders(stocks_list, total_value / len(stocks_list))




# 批量买卖股票
def bulk_orders(stocks_list, target_value):
    for i in stocks_list                      :
        order_target_value(i, target_value, pindex=0)


def bete_order(context, money):
    amount = int( money / get_current_data()[g.betaName].last_price)
    order_target(g.betaName, -amount , side='long', pindex=1)  # 做空

'''
END 调仓
'''

'''
调仓初版
'''


##获得卖出信号，并执行卖出操作
# 输入：context, data，已排序股票列表stock_sort-list类型
# 输出：none
def order_stock_sell(context, data, stock_sort):
    # 对于不需要持仓的股票，全仓卖出
    for stock in context.portfolio.positions:
        # 除去排名前g.N个股票（选股！）
        if stock not in stock_sort:
            stock_sell = stock
            order_target_value(stock_sell, 0)


# 获得买入信号，并执行买入操作
# 输入：context, data，已排序股票列表stock_sort-list类型
# 输出：none
def order_stock_buy(context, data, stock_sort):
    # 对于需要持仓的股票，按分配到的份额买入
    for stock in stock_sort:
        stock_buy = stock
        order_target_value(stock_buy, g.everyStock)


'''
END 调仓初版
'''