Question

我一直在尝试使用QuantLib（Python）建立债券曲线，但是我获得的合理数量的证券的拟合度确实很差（直线）。也许我需要校准参数（simplexLambda或迭代次数或精度）以使其起作用，但是我还没有找到解决方案。

我已经包含了创建一些债券，然后使用SvenssonFitting（）和FittedBondDiscountCurve（）生成收益曲线的代码。

import QuantLib as ql
from matplotlib import pyplot as plt


ql_date = ql.Date(1,7,2019)


BOND_METADATA = [[99.40625, '2017-07-17', '2020-07-15', 0.015],
                [101.4101563, '2018-07-16', '2021-07-15', 0.02625],
                [99.59765625, '2017-06-30', '2022-06-30', 0.0175],
                [102.7695313, '2018-07-02', '2023-06-30', 0.02625],
                [100.2695313, '2017-06-30', '2024-06-30', 0.02],
                [104.265625, '2018-07-02', '2025-06-30', 0.0275],
                [96.3125, '2016-08-15', '2026-08-15', 0.015],
                [101.171875, '2017-08-15', '2027-08-15', 0.0225],
                [106.1328125, '2018-08-15', '2028-08-15', 0.02875],
                [108.3125, '2018-11-15', '2028-11-15', 0.03125],
                [116.9609375, '2009-02-17', '2039-02-15', 0.035],
                [133.1171875, '2009-08-17', '2039-08-15', 0.045],
                [123, '2010-08-16', '2040-08-15', 0.03875],
                [120.8984375, '2011-08-15', '2041-08-15', 0.0375],
                [103.2265625, '2012-08-15', '2042-08-15', 0.0275],
                [109.71875, '2014-08-15', '2044-08-15', 0.03125],
                [92.7265625, '2016-08-15', '2046-08-15', 0.0225],
                [107.546875, '2018-08-15', '2048-08-15', 0.03],
                [115.46875, '2018-11-15', '2048-11-15', 0.03375]]


bond_helpers = []

for i,(price,issue_date,maturity_date,coupon) in enumerate(BOND_METADATA):
    face_value = 100.
    settlement_days = 1
    issue_date = ql.Date(issue_date,"%Y-%m-%d")
    maturity_date = ql.Date(maturity_date,"%Y-%m-%d")
    schedule = ql.Schedule(issue_date,
                           maturity_date,
                           ql.Period(6,ql.Months),
                           ql.UnitedStates(),
                           ql.Unadjusted,
                           ql.Unadjusted,
                           ql.DateGeneration.Backward,
                           False)
    security = ql.FixedRateBond(settlement_days,
                                   face_value,
                                   schedule,
                                   [coupon],
                                   ql.ActualActual()
                                   )
    bond_helpers.append(ql.FixedRateBondHelper(ql.QuoteHandle(ql.SimpleQuote(price)),
                                       security.settlementDays(),
                                       security.notional(),
                                       schedule,
                                       [security.nextCouponRate()],
                                       security.dayCounter(),
                                       ql.Following))


guess = []
simplexLambda = 5
accuracy = 1e-8
numIter = 100000
method = ql.SvenssonFitting()
yield_curve = ql.FittedBondDiscountCurve(ql_date,
                                         bond_helpers,
                                         ql.ActualActual(),
                                         method,
                                         accuracy,
                                         numIter,
                                         guess,
                                         5)
#---------------------------------------------------------

t0 = yield_curve.referenceDate()
x = [(t0+i).to_date() for i in range(2500)]
y = [yield_curve.forwardRate(t0+i,t0+i+1,ql.ActualActual(),ql.Compounded,ql.Semiannual).rate() for i in range(2500)]
plt.plot(x,y,color='orange')

我得到的拟合看起来像一条直线，它不应该给出输入的产量。可以将图片发布为支持文档。

A bond curve fit

Quantlib Svensson拟合问题

0 个答案: