 ROOT   Reference Guide
Searching...
No Matches
portfolio.C File Reference

## Detailed Description  This macro shows in detail the use of the quadratic programming package quadp .

Running this macro :

.x portfolio.C+

or

.L portFolio.C+; portfolio()
R__EXTERN TSystem * gSystem
Definition TSystem.h:560
virtual int Load(const char *module, const char *entry="", Bool_t system=kFALSE)
Definition TSystem.cxx:1842

Let's first review what we exactly mean by "quadratic programming" :

We want to minimize the following objective function :

$$c^T x + ( 1/2 ) x^T Q x$$ wrt. the vector $$x$$

$$c$$ is a vector and $$Q$$ a symmetric positive definite matrix

You might wonder what is so special about this objective which is quadratic in the unknowns, that can not be done by Minuit/Fumili . Well, we have in addition the following boundary conditions on $$x$$:

$A x = b \\ clo \le C x \le cup \\ xlo \le x \le xup$

where A and C are arbitrary matrices and the rest are vectors

Not all these constraints have to be defined . Our example will only use $$xlo$$, $$A$$ and $$b$$ Still, this could be handled by a general non-linear minimizer like Minuit by introducing so-called "slack" variables . However, quadp is tailored to objective functions not more complex than being quadratic . This allows usage of solving techniques which are even stable for problems involving for instance 500 variables, 100 inequality conditions and 50 equality conditions .

Enough said about quadratic programming, let's return to our example . Suppose, after a long day of doing physics, you have a look at your investments and realize that an early retirement is not possible, given the returns of your stocks . So what now ? ROOT to the rescue ...

In 1990 Harry Markowitz was awarded the Nobel prize for economics: " his work provided new tools for weighing the risks and rewards of different investments and for valuing corporate stocks and bonds" . In plain English, he developed the tools to balance greed and fear, we want the maximum return with the minimum amount of risk. Our stock portfolio should be at the "Efficient Frontier". To quantify better the risk we are willing to take, we define a utility function $$U(x)$$. It describes as a function of our total assets $$x$$, our "satisfaction" . A common choice is $$1-exp(-k*x)$$ (the reason for the exponent will be clear later) . The parameter $$k$$ is the risk-aversion factor . For small values of $$k$$ the satisfaction is small for small values of $$x$$; by increasing $$x$$ the satisfaction can still be increased significantly . For large values of $$k$$, $$U(x)$$ increases rapidly to 1, there is no increase in satisfaction for additional dollars earned .

In summary :

• small $$k$$ ==> risk-loving investor
• large $$k$$ ==> risk-averse investor

Suppose we have for nrStocks the historical daily returns $$r = closing_price(n) - closing_price(n-1)$$. Define a vector $$x$$ of length of $$nrStocks$$, which contains the fraction of our money invested in each stock . We can calculate the average daily return $$z$$ of our portfolio and its variance using the portfolio covariance Covar :

$$z = r^T x$$ and $$var = x^T Covar x$$

Assuming that the daily returns have a Normal distribution, $$N(x)$$, so will $$z$$ with mean $$r^T x$$ and variance $$x^T Covar x$$

The expected value of the utility function is :

$E(u(x)) = Int (1-exp(-k*x) N(x) dx \\ = 1-exp(-k (r^T x - 0.5 k x^T Covar x) ) \\$

Its value is maximised by maximising $$r^T x -0.5 k x^T Covar x$$ under the condition $$sum (x_i) = 1$$, meaning we want all our money invested and $$x_i \ge 0$$, we can not "short" a stock

For 10 stocks we got the historical daily data for Sep-2000 to Jun-2004:

• GE : General Electric Co
• SUNW : Sun Microsystems Inc
• QCOM : Qualcomm Inc
• TYC : Tyco International Ltd
• IBM : International Business Machines Corp
• AMAT : Applied Materials Inc
• C : Citigroup Inc
• PFE : Pfizer Inc
• HD : Home Depot Inc

We calculate the optimal portfolio for 2.0 and 10.0 .

Food for thought :

• We assumed that the stock returns have a Normal distribution . Check this assumption by histogramming the stock returns !
• We used for the expected return in the objective function, the flat average over a time period . Investment firms will put significant resources in improving the return prediction .
• If you want to trade significant number of shares, several other considerations have to be taken into account :
• If you are going to buy, you will drive the price up (so-called "slippage") . This can be taken into account by adding terms to the objective (Google for "slippage optimization")
• FTC regulations might have to be added to the inequality constraints
• Investment firms do not want to be exposed to the "market" as defined by a broad index like the S&P and "hedge" this exposure away . A perfect hedge this can be added as an equality constrain, otherwise add an inequality constrain . stock daily daily w1 w2
symb return sdv
GE : 1.001 0.022 0.000 0.134
SUNW : 1.004 0.047 0.676 0.145
QCOM : 1.001 0.039 0.000 0.000
BRCM : 1.003 0.056 0.179 0.035
TYC : 1.001 0.042 0.145 0.069
IBM : 1.001 0.023 0.000 0.096
AMAT : 1.001 0.040 0.000 0.000
C : 1.000 0.023 0.000 0.000
PFE : 1.000 0.019 0.000 0.424
HD : 1.001 0.029 0.000 0.098
#include "Riostream.h"
#include "TCanvas.h"
#include "TFile.h"
#include "TMath.h"
#include "TTree.h"
#include "TArrayF.h"
#include "TH1.h"
#include "TF1.h"
#include "TLegend.h"
#include "TSystem.h"
#include "TMatrixD.h"
#include "TMatrixDSym.h"
#include "TVectorD.h"
#include "TQpProbDens.h"
#include "TGondzioSolver.h"
const Int_t nrStocks = 10;
static const Char_t *stocks[] =
{"GE","SUNW","QCOM","BRCM","TYC","IBM","AMAT","C","PFE","HD"};
class TStockDaily {
public:
Int_t fDate;
Int_t fOpen; // 100*open_price
Int_t fHigh; // 100*high_price
Int_t fLow; // 100*low_price
Int_t fClose; // 100*close_price
Int_t fVol;
TStockDaily() {
fDate = fVol = fOpen = fHigh = fLow = fClose = fCloseAdj = 0;
}
virtual ~TStockDaily() {}
ClassDef(TStockDaily,1)
};
//---------------------------------------------------------------------------
Double_t RiskProfile(Double_t *x, Double_t *par) {
Double_t riskFactor = par;
return 1-TMath::Exp(-riskFactor*x);
}
//---------------------------------------------------------------------------
TArrayF &StockReturn(TFile *f,const TString &name,Int_t sDay,Int_t eDay)
{
TTree *tDaily = (TTree*)f->Get(name);
TStockDaily *data = 0;
TBranch *b_date = tDaily->GetBranch("fDate");
const Int_t nrEntries = (Int_t)tDaily->GetEntries();
for (Int_t i = 0; i < nrEntries; i++) {
b_date->GetEntry(i);
if (data->fDate >= sDay && data->fDate <= eDay)
}
TArrayF *r = new TArrayF(nrEntries-1);
for (Int_t i = 1; i < nrEntries; i++)
return *r;
}
#ifndef __MAKECINT__
//---------------------------------------------------------------------------
TVectorD OptimalInvest(Double_t riskFactor,TVectorD r,TMatrixDSym Covar)
{
// what the quadratic programming package will do:
//
// minimize c^T x + ( 1/2 ) x^T Q x
// subject to A x = b
// clo <= C x <= cup
// xlo <= x <= xup
// what we want :
//
// maximize c^T x - k ( 1/2 ) x^T Q x
// subject to sum_x x_i = 1
// 0 <= x_i
// We have nrStocks weights to determine,
// 1 equality- and 0 inequality- equations (the simple square boundary
// condition (xlo <= x <= xup) does not count)
const Int_t nrVar = nrStocks;
const Int_t nrEqual = 1;
const Int_t nrInEqual = 0;
// flip the sign of the objective function because we want to maximize
TVectorD c = -1.*r;
TMatrixDSym Q = riskFactor*Covar;
// equality equation
TMatrixD A(nrEqual,nrVar); A = 1;
TVectorD b(nrEqual); b = 1;
// inequality equation
//
// - although not applicable in the current situation since nrInEqual = 0, one
// has to specify not only clo and cup but also an index vector iclo and icup,
// whose values are either 0 or 1 . If iclo[j] = 1, the lower boundary condition
// is active on x[j], etc. ...
TMatrixD C (nrInEqual,nrVar);
TVectorD clo (nrInEqual);
TVectorD cup (nrInEqual);
TVectorD iclo(nrInEqual);
TVectorD icup(nrInEqual);
// simple square boundary condition : 0 <= x_i, so only xlo is relevant .
// Like for clo and cup above, we have to define an index vector ixlo and ixup .
// Since each variable has the lower boundary, we can set the whole vector
// ixlo = 1
TVectorD xlo (nrVar); xlo = 0;
TVectorD xup (nrVar); xup = 0;
TVectorD ixlo(nrVar); ixlo = 1;
TVectorD ixup(nrVar); ixup = 0;
// setup the quadratic programming problem . Since a small number of variables are
// involved and "Q" has everywhere entries, we chose the dense version "TQpProbDens" .
// In case of a sparse formulation, simply replace all "Dens" by "Sparse" below and
// use TMatrixDSparse instead of TMatrixDSym and TMatrixD
TQpProbDens *qp = new TQpProbDens(nrVar,nrEqual,nrInEqual);
// stuff all the matrices/vectors defined above in the proper places
// setup the nrStock variables, vars->fX will contain the final solution
TQpVar *vars = qp->MakeVariables(prob);
TQpResidual *resid = qp->MakeResiduals(prob);
// Now we have to choose the method of solving, either TGondzioSolver or TMehrotraSolver
// The Gondzio method is more sophisticated and therefore numerically more involved
// If one want the Mehrotra method, simply replace "Gondzio" by "Mehrotra" .
TGondzioSolver *s = new TGondzioSolver(qp,prob);
const Int_t status = s->Solve(prob,vars,resid);
const TVectorD weight = vars->fX;
delete qp; delete prob; delete vars; delete resid; delete s;
if (status != 0) {
cout << "Could not solve this problem." <<endl;
return TVectorD(nrStocks);
}
return weight;
}
#endif
//---------------------------------------------------------------------------
void portfolio()
{
const Int_t sDay = 20000809;
const Int_t eDay = 20040602;
const char *fname = "stock.root";
TFile *f = 0;
if (!gSystem->AccessPathName(fname)) {
f = TFile::Open(fname);
} else if (!gSystem->AccessPathName(Form("%s/quadp/%s", TROOT::GetTutorialDir().Data(), fname))) {
} else {
printf("accessing %s file from http://root.cern.ch/files\n",fname);
f = TFile::Open(Form("http://root.cern.ch/files/%s",fname));
}
if (!f) return;
TArrayF *data = new TArrayF[nrStocks];
for (Int_t i = 0; i < nrStocks; i++) {
const TString symbol = stocks[i];
data[i] = StockReturn(f,symbol,sDay,eDay);
}
const Int_t nrData = data.GetSize();
TVectorD r(nrStocks);
for (Int_t i = 0; i < nrStocks; i++)
r[i] = data[i].GetSum()/nrData;
TMatrixDSym Covar(nrStocks);
for (Int_t i = 0; i < nrStocks; i++) {
for (Int_t j = 0; j <= i; j++) {
Double_t sum = 0.;
for (Int_t k = 0; k < nrData; k++) {
sum += (data[i][k] - r[i]) * (data[j][k] - r[j]);
}
Covar(i,j) = Covar(j,i) = sum/nrData;
}
}
const TVectorD weight1 = OptimalInvest(2.0,r,Covar);
const TVectorD weight2 = OptimalInvest(10.,r,Covar);
cout << "stock daily daily w1 w2" <<endl;
cout << "symb return sdv " <<endl;
for (Int_t i = 0; i < nrStocks; i++)
printf("%s\t: %.3f %.3f %.3f %.3f\n",stocks[i],r[i],TMath::Sqrt(Covar[i][i]),weight1[i],weight2[i]);
TCanvas *c1 = new TCanvas("c1","Portfolio Optimizations",10,10,800,900);
c1->Divide(1,2);
// utility function / risk profile
c1->cd(1);
TF1 *f1 = new TF1("f1",RiskProfile,0,2.5,1);
f1->SetParameter(0,2.0);
f1->Draw("AC");
f1->GetHistogram()->SetXTitle("dollar");
f1->GetHistogram()->SetYTitle("utility");
TF1 *f2 = new TF1("f2",RiskProfile,0,2.5,1);
f2->SetParameter(0,10.);
f2->SetLineColor(50);
f2->Draw("CSAME");
TLegend *legend1 = new TLegend(0.50,0.65,0.70,0.82);
legend1->Draw();
// vertical bar chart of portfolio distribution
c1->cd(2);
TH1F *h1 = new TH1F("h1","Portfolio Distribution",nrStocks,0,0);
TH1F *h2 = new TH1F("h2","Portfolio Distribution",nrStocks,0,0);
h1->SetStats(0);
h2->SetFillColor(50);
h1->SetBarWidth(0.45);
h2->SetBarWidth(0.4);
h2->SetBarOffset(0.55);
for (Int_t i = 0; i < nrStocks; i++) {
h1->Fill(stocks[i],weight1[i]);
h2->Fill(stocks[i],weight2[i]);
}
h1->Draw("BAR2 HIST");
h2->Draw("BAR2SAME HIST");
TLegend *legend2 = new TLegend(0.50,0.65,0.70,0.82);
legend2->Draw();
}
#define b(i)
Definition RSha256.hxx:100
#define f(i)
Definition RSha256.hxx:104
#define c(i)
Definition RSha256.hxx:101
int Int_t
Definition RtypesCore.h:45
char Char_t
Definition RtypesCore.h:37
double Double_t
Definition RtypesCore.h:59
#define ClassDef(name, id)
Definition Rtypes.h:337
Option_t Option_t TPoint TPoint const char GetTextMagnitude GetFillStyle GetLineColor GetLineWidth GetMarkerStyle GetTextAlign GetTextColor GetTextSize void data
Option_t Option_t TPoint TPoint const char GetTextMagnitude GetFillStyle GetLineColor GetLineWidth GetMarkerStyle GetTextAlign GetTextColor GetTextSize void char Point_t Rectangle_t WindowAttributes_t Float_t r
char name
Definition TGX11.cxx:110
char * Form(const char *fmt,...)
Formats a string in a circular formatting buffer.
Definition TString.cxx:2468
TVectorT< Double_t > TVectorD
Definition TVectorDfwd.h:23
Array of floats (32 bits per element).
Definition TArrayF.h:27
virtual void SetFillColor(Color_t fcolor)
Set the fill area color.
Definition TAttFill.h:37
virtual void SetLineColor(Color_t lcolor)
Set the line color.
Definition TAttLine.h:40
A TTree is a list of TBranches.
Definition TBranch.h:93
virtual Int_t GetEntry(Long64_t entry=0, Int_t getall=0)
Definition TBranch.cxx:1706
The Canvas class.
Definition TCanvas.h:23
1-Dim function class
Definition TF1.h:213
virtual TH1 * GetHistogram() const
Return a pointer to the histogram used to visualise the function Note that this histogram is managed ...
Definition TF1.cxx:1585
void Draw(Option_t *option="") override
Draw this function with its current attributes.
Definition TF1.cxx:1334
virtual void SetParameter(Int_t param, Double_t value)
Definition TF1.h:639
A ROOT file is composed of a header, followed by consecutive data records (TKey instances) with a wel...
Definition TFile.h:54
static TFile * Open(const char *name, Option_t *option="", const char *ftitle="", Int_t compress=ROOT::RCompressionSetting::EDefaults::kUseCompiledDefault, Int_t netopt=0)
Create / open a file.
Definition TFile.cxx:4070
Derived class of TQpSolverBase implementing Gondzio-correction version of Mehrotra's original predict...
Int_t Solve(TQpDataBase *prob, TQpVar *iterate, TQpResidual *resid) override
Solve the quadratic programming problem as formulated through prob, store the final solution in itera...
1-D histogram with a float per channel (see TH1 documentation)}
Definition TH1.h:577
virtual void SetBarOffset(Float_t offset=0.25)
Set the bar offset as fraction of the bin width for drawing mode "B".
Definition TH1.h:361
virtual void SetXTitle(const char *title)
Definition TH1.h:415
virtual void SetMaximum(Double_t maximum=-1111)
Definition TH1.h:400
virtual Int_t Fill(Double_t x)
Increment bin with abscissa X by 1.
Definition TH1.cxx:3341
void Draw(Option_t *option="") override
Draw this histogram with options.
Definition TH1.cxx:3063
virtual void SetMinimum(Double_t minimum=-1111)
Definition TH1.h:401
virtual void SetYTitle(const char *title)
Definition TH1.h:416
virtual void SetBarWidth(Float_t width=0.5)
Set the width of bars as fraction of the bin width for drawing mode "B".
Definition TH1.h:362
virtual void SetStats(Bool_t stats=kTRUE)
Set statistics option on/off.
Definition TH1.cxx:8859
This class displays a legend box (TPaveText) containing several legend entries.
Definition TLegend.h:23
TLegendEntry * AddEntry(const TObject *obj, const char *label="", Option_t *option="lpf")
Add a new entry to this legend.
Definition TLegend.cxx:317
void Draw(Option_t *option="") override
Draw this legend with its current attributes.
Definition TLegend.cxx:422
Data for the dense QP formulation.
dense matrix problem formulation
Definition TQpProbDens.h:61
TQpVar * MakeVariables(const TQpDataBase *data) override
Setup the variables.
virtual TQpDataBase * MakeData(Double_t *c, Double_t *Q, Double_t *xlo, Bool_t *ixlo, Double_t *xup, Bool_t *ixup, Double_t *A, Double_t *bA, Double_t *C, Double_t *clo, Bool_t *iclo, Double_t *cup, Bool_t *icup)
Setup the data.
TQpResidual * MakeResiduals(const TQpDataBase *data) override
Setup the residuals.
The Residuals class calculates and stores the quantities that appear on the right-hand side of the li...
Definition TQpResidual.h:62
Class containing the variables for the general QP formulation.
Definition TQpVar.h:60
TVectorD fX
Definition TQpVar.h:91
static const TString & GetTutorialDir()
Get the tutorials directory in the installation. Static utility function.
Definition TROOT.cxx:3064
Basic string class.
Definition TString.h:139
virtual Bool_t AccessPathName(const char *path, EAccessMode mode=kFileExists)
Returns FALSE if one can access a file using the specified access mode.
Definition TSystem.cxx:1283
A TTree represents a columnar dataset.
Definition TTree.h:79
virtual TBranch * GetBranch(const char *name)
Return pointer to the branch with the given name in this tree or its friends.
Definition TTree.cxx:5291
Change branch address, dealing with clone trees properly.
Definition TTree.cxx:8377
virtual Long64_t GetEntries() const
Definition TTree.h:463
return c1
Definition legend1.C:41
Double_t x[n]
Definition legend1.C:17
TH1F * h1
Definition legend1.C:5
TF1 * f1
Definition legend1.C:11
constexpr Double_t C()
Velocity of light in .
Definition TMath.h:114
Double_t Exp(Double_t x)
Returns the base-e exponential function of x, which is e raised to the power x.
Definition TMath.h:709
Double_t Sqrt(Double_t x)
Returns the square root of x.
Definition TMath.h:662
static uint64_t sum(uint64_t i)
Definition Factory.cxx:2345

Definition in file portfolio.C.