user-defined package

Const Variable

获取数据，清洗数据

```py
def data_for_acb(ticker="000001", tstart=2010, tend=2015):
    """获取，清洗 ACB 模型所需要的数据。

    ACB 模型需要数据：
        负债合计［TLiab］
        资产总计［TAssets］
        未分配利润［retainedEarnings］
        净利润［NIncome］
        营运资本 = 资产总计 - 负债合计
    """
    bs_data = DataAPI.FdmtBSGet(ticker=ticker, beginYear=tstart-1, endYear=tend, 
                                field=['secID', 'endDate', 'publishDate', 'TLiab', 'TAssets', 'retainedEarnings'])
    is_data = DataAPI.FdmtISGet(ticker=ticker, beginYear=tstart-1, endYear=tend, 
                                field=['secID', 'endDate', 'publishDate', 'NIncome'])
    bs_data = bs_data.drop_duplicates('endDate')
    is_data = is_data.drop_duplicates('endDate')

    data = is_data.merge(bs_data, on=['secID', 'endDate'])

    # calculate TAssets diff of current and last report
    pre_TAssets = []
    length = len(data)

    for index, number in enumerate(data.TAssets):
        if index + 1 == length:
            last_number = index
        else:
            last_number = index + 1
        pre_TAssets.append(data.TAssets[last_number])

    data['TAssetsPre'] = pre_TAssets

    return data

def data_for_acbel(ticker="000001", tstart=2010, tend=2015):
    """获取，清洗 ACB 模型所需要的数据。

    ACB 模型需要数据：
        负债合计［TLiab］
        资产总计［TAssets］
        未分配利润［retainedEarnings］
        净利润［NIncome］
        营业总收入［tRevenue］
        总市值［marketValue］
    """
    bs_data = DataAPI.FdmtBSGet(ticker=ticker, beginYear=tstart-1, endYear=tend, 
                                field=['secID', 'endDate', 'publishDate', 'TLiab', 'TAssets', 'retainedEarnings'])
    is_data = DataAPI.FdmtISGet(ticker=ticker, beginYear=tstart-1, endYear=tend, 
                                field=['secID', 'endDate', 'publishDate', 'NIncome', 'tRevenue'])
    market_data = DataAPI.MktEqudGet(ticker=ticker,  field=['secID', 'tradeDate', 'marketValue'])
    market_data.rename(columns={'tradeDate': 'endDate'}, inplace=True)
    bs_data = bs_data.drop_duplicates('endDate')
    is_data = is_data.drop_duplicates('endDate')

    data = is_data.merge(bs_data, on=['secID', 'endDate'])
    endDate = list(data.endDate)
    data = data.merge(market_data, on=['secID', 'endDate'], how='outer')
    data.marketValue = data.marketValue.fillna(method='ffill')
    data = data[data.endDate.isin(endDate)]
    # calculate TAssets diff of current and last report
    pre_TAssets = []
    length = len(data)

    for index, number in enumerate(data.TAssets):
        if index + 1 == length:
            last_number = index
        else:
            last_number = index + 1
        pre_TAssets.append(data.TAssets[last_number])

    data['TAssetsPre'] = pre_TAssets

    return data

测试：获取数据，清洗数据

data_for_acb("002056").head(5)

data_for_acbel("002056").head(5)

计算 Z-score

def zscore_ACB(ticker=None, tstart=2010, tend=2015, coef=[0.517, -0.460, 18.640, 0.388, 1.158]):
    # step 1. get data and pre-calculate the factor
    ticker = data_for_acb(ticker, tstart, tend)
    ticker['x0'] = 1
    ticker['x1'] = ticker['TLiab'] / ticker['TAssets']
    ticker['x2'] = ticker['NIncome'] * 2 / (ticker['TAssets'] + ticker['TAssetsPre'])
    ticker['x3'] = (ticker['TAssets'] - ticker['TLiab']) / ticker['TAssets']
    ticker['x4'] = ticker['retainedEarnings'] / ticker['TAssets']

    # step 2. calculate zscore 
    tmp = ticker[['x0', 'x1', 'x2', 'x3', 'x4']] * coef
    ticker['zscore'] = tmp.sum(axis=1)

    # step 3. build result
    ticker.sort('endDate', ascending=True, inplace=True)
    return ticker[['secID', 'endDate', 'NIncome', 'TLiab', 'TAssets', 
                   'retainedEarnings', 'x0', 'x1', 'x2', 'x3', 'x4', 'zscore']]


def zscore_ACBEL(ticker=None, tstart=2010, tend=2015, coef=[0.2086, 4.3465, 4.9601]):
    # step 1. get data and pre-calculate the factor
    ticker = data_for_acbel(ticker, tstart, tend)
    ticker['x0'] = ticker['marketValue'] / ticker['TLiab']
    ticker['x1'] = ticker['tRevenue'] / ticker['TAssets']
    ticker['x2'] = (ticker['TAssets'] - ticker['TAssetsPre']) / ticker['TAssetsPre']

    # step 2. calculate zscore 
    tmp = ticker[['x0', 'x1', 'x2']] * coef
    ticker['zscore'] = tmp.sum(axis=1)

    # step 3. build result
    ticker.sort('endDate', ascending=True, inplace=True)
    return ticker[['secID', 'endDate', 'NIncome', 'TLiab', 'tRevenue', 'TAssets', 
                   'retainedEarnings', 'marketValue', 'TAssetsPre', 'x0', 'x1', 'x2', 'zscore']]

def get_ticker(bond=None):
    """Get the ticker number of a bond.
    """
    # bondID -> partyID -> ticker
    partyID = None
    try:
        data = DataAPI.BondGet(ticker=bond)
        partyID = data['partyID'][0]
    except:
        return 'Cannot find this bond in DataAPI'

    ticker = None
    try:
        data = DataAPI.SecIDGet(partyID=str(partyID))
        ticker = data['ticker'][0]
    except:
        return 'Cannot find the ticker for this bond in DataAPI, maybe the issuer is not listed'

    return ticker

测试：计算 Z-score

zscore_ACB("002506").head(5)

zscore_ACBEL("002506").head(5)

a = zscore_ACBEL("002506")
a.head(3)

作图分析

def zscore_plot(dataframe, upper_limit, low_limit):    
    ax = dataframe.plot('endDate', ['zscore'], figsize=(20, 10), style='g-', title='zscore curve',)
    axhspan(low_limit, dataframe.zscore.min(), facecolor='maroon', alpha=0.1)
    axhspan(upper_limit, dataframe.zscore.max(), facecolor='yellow', alpha=0.2)
    ax.legend()
    return ax

测试：作图分析

这里，我们以 11超日债［112061］ 来做测试，看看当前这个模型表现怎么样。

• step 1: 调用 get_ticker 函数通过债券代码获取发行人上市代码，在发行人已上市的前提下；
• step 2: 调用 zscore_ACBEL 或 zscore_ACB 计算发行人的 Z-score 值；
• step 3: 调用 zscore_plot 绘制 Z-score 曲线；

ticker = get_ticker("112061")
df = zscore_ACBEL(ticker)
zscore_plot(df, 1.5408, 1.5408)

<matplotlib.axes.AxesSubplot at 0x5220a10>

ticker = get_ticker("112061")
df = zscore_ACB(ticker)
zscore_plot(df, 0.9, 0.5)

<matplotlib.axes.AxesSubplot at 0x525c610>