Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
braverock
GitHub Repository: braverock/portfolioanalytics
 Path: blob/master/sandbox/mult_layer_script.R
1433 views
1


2
data(edhec)

3
R <- edhec[, 1:10]

4
funds <- colnames(R)

5


6
# The first sub-portfolio, portf1, will contain assets 1:5 of edhec

7
# with an objective to minimize standard deviation.

8
portf1 <- portfolio.spec(assets=funds[1:5])

9
portf1 <- add.constraint(portfolio=portf1, type="weight_sum", 

10
                         min_sum=0.99, max_sum=1.01)

11
portf1 <- add.constraint(portfolio=portf1, type="long_only")

12
portf1 <- add.objective(portfolio=portf1, type="risk", name="StdDev")

13


14
# The second sub-portfolio, portf2, will contain assets 6:10 of edhec

15
# with an objective to minimize expected shortfall.

16
portf2 <- portfolio.spec(assets=funds[6:10])

17
# portf2 <- portfolio.spec(assets=5)

18
portf2 <- add.constraint(portfolio=portf2, type="weight_sum", 

19
                         min_sum=0.99, max_sum=1.01)

20
portf2 <- add.constraint(portfolio=portf2, type="long_only")

21
portf2 <- add.objective(portfolio=portf2, type="risk", name="ES",

22
                            arguments=list(p=0.9))

23


24
# portf1 and portf2 have the same constraints so they can used the same

25
# set of random portfolios

26
set.seed(123)

27
rp <- random_portfolios(portf2, 2000)

28


29


30
# The 'top level' portfolio has objectives for equal contribution to risk

31
# where modified ES is the risk measure

32
portf <- portfolio.spec(assets=paste("proxy",1:2, sep="."))

33
portf <- add.constraint(portfolio=portf, type="weight_sum", 

34
                        min_sum=0.99, max_sum=1.01)

35
portf <- add.constraint(portfolio=portf, type="long_only")

36
portf <- add.objective(portfolio=portf, type="risk", name="ES",

37
                        arguments=list(p=0.9))

38
portf <- add.objective(portfolio=portf, type="risk_budget", name="ES",

39
                       arguments=list(p=0.9), min_concentration=TRUE)

40


41
# Specify a mult-layer portfolio

42
mult.portf <- mult.portfolio.spec(portf)

43


44
# Add portf1 as a sub portfolio with optimization parameters specific to 

45
# running optimize.portfolio.rebalancing with portf1

46
mult.portf <- add.sub.portfolio(mult.portf, portf1, rp=rp, 

47
                                optimize_method="random",

48
                                rebalance_on="quarters", 

49
                                training_period=136)

50


51
# Add portf2 as a sub portfolio with optimization parameters specific to 

52
# running optimize.portfolio.rebalancing with portf2

53
mult.portf <- add.sub.portfolio(mult.portf, portf2, rp=rp,

54
                                optimize_method="random",

55
                                rebalance_on="months", 

56
                                training_period=136, 

57
                                trailing_periods=48)

58


59
# Compute the out of sample backtesting returns for each sub portfolio

60
proxy.ret <- PortfolioAnalytics:::proxy.mult.portfolio(R, mult.portf)

61


62
# Verify that proxy.mult.portfolio is computing returns correctly

63
opt1 <- optimize.portfolio.rebalancing(R[,1:5], 

64
                                       portf1, 

65
                                       optimize_method="random",

66
                                       rp=rp,

67
                                       rebalance_on="quarters", 

68
                                       training_period=136)

69
ret1 <- summary(opt1)$portfolio_returns

70


71
opt2 <- optimize.portfolio.rebalancing(R[,6:10], 

72
                                       portf2, 

73
                                       optimize_method="random",

74
                                       rp=rp,

75
                                       rebalance_on="months", 

76
                                       training_period=136,

77
                                       trailing_periods=48)

78
ret2 <- summary(opt2)$portfolio_returns

79
ret <- na.omit(cbind(ret1, ret2))

80


81
all.equal(ret, proxy.ret, check.attributes=FALSE)

82


83
# Verify that multi layer optimization is done correctly in optimize.portfolio

84
set.seed(123)

85
rp.top <- random_portfolios(portf, 1000)

86


87
opt <- optimize.portfolio(proxy.ret, portf, 

88
                          optimize_method="random", 

89
                          trace=TRUE, rp=rp.top)

90


91
opt.mult <- optimize.portfolio(R, mult.portf,

92
                               optimize_method="random", 

93
                               trace=TRUE, rp=rp.top)

94


95
all.equal(extractObjectiveMeasures(opt), extractObjectiveMeasures(opt.mult))

96
all.equal(extractWeights(opt), extractWeights(opt.mult))

97


98