pith. sign in

arxiv: 1907.07101 · v1 · pith:SH3B3HUVnew · submitted 2019-07-15 · 💱 q-fin.PM

Location and portfolio selection problems: A unified framework

Justo Puerto , Moises Rodr\'iguez-Madrena , Andrea Scozzari This is my paper

Pith reviewed 2026-05-24 21:20 UTC · model grok-4.3

classification 💱 q-fin.PM
keywords portfolio selectionlocation problemsmixed-integer linear programmingclusteringassets networkcorrelation metricunified optimization
0
0 comments X

The pith

A mixed-integer linear program solves asset clustering and portfolio selection together instead of in separate steps.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes combining classical location problems on networks with portfolio optimization into a single optimization model. By using correlation coefficients to define distances between assets, the model treats clustering as an additional criterion in the portfolio problem. This unified approach avoids the common two-step process of first clustering assets and then selecting the portfolio. A mixed-integer linear programming formulation makes this possible, and preliminary tests on real financial data show its effectiveness. If correct, this means investors can optimize for both return-risk and clustering effects in one computation.

Core claim

The paper shows that by endowing the assets network with a correlation-based metric and adding a classical location problem objective as a new criterion to the portfolio selection problem, a mixed-integer linear programming formulation can handle the combined clustering and selection in a unified phase.

What carries the argument

A mixed-integer linear programming formulation that integrates an objective function from a classical location problem into the mean-risk portfolio optimization model.

If this is right

  • The model measures the effect of clustering on selected assets versus non-selected ones directly.
  • It allows solving the portfolio problem while accounting for network clustering in one optimization run.
  • Preliminary computational experiments on real datasets validate its effectiveness.
  • This approach can be applied to mean-risk bi-criteria optimization with clustering.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This could lead to more stable portfolios by considering asset clusters implicitly.
  • Extensions might include other location problems like p-median or facility location variants.
  • It suggests that correlation networks can serve as a basis for joint financial and spatial optimization problems.

Load-bearing premise

The assumption that adding a location problem objective function as a criterion meaningfully captures the clustering effect on selected assets relative to non-selected ones, using correlation as distance.

What would settle it

If running the unified MILP on a dataset where clustering is known to matter shows no improvement in out-of-sample performance compared to separate clustering then selection, the unified benefit would be refuted.

read the original abstract

Given a set of assets and an investment capital, the classical portfolio selection problem consists in determining the amount of capital to be invested in each asset in order to build the most profitable portfolio. The portfolio optimization problem is naturally modeled as a mean-risk bi-criteria optimization problem where the mean rate of return of the portfolio must be maximized whereas a given risk measure must be minimized. Several mathematical programming models and techniques have been presented in the literature in order to efficiently solve the portfolio problem. A relatively recent promising line of research is to exploit clustering information of an assets network in order to develop new portfolio optimization paradigms. In this paper we endow the assets network with a metric based on correlation coefficients between assets' returns, and show how classical location problems on networks can be used for clustering assets. In particular, by adding a new criterion to the portfolio selection problem based on an objective function of a classical location problem, we are able to measure the effect of clustering on the selected assets with respect to the non-selected ones. Most papers dealing with clustering and portfolio selection models solve these problems in two distinct steps: cluster first and then selection. The innovative contribution of this paper is that we propose a Mixed-Integer Linear Programming formulation for dealing with this problem in a unified phase. The effectiveness of our approach is validated reporting some preliminary computational experiments on some real financial dataset.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes a mixed-integer linear programming (MILP) formulation that unifies mean-risk portfolio optimization with a classical location problem objective defined on an assets network whose edges are weighted by a correlation-based metric; this allows simultaneous asset clustering and portfolio selection in a single optimization phase rather than the conventional two-step cluster-then-select procedure, with preliminary computational results reported on real financial datasets.

Significance. If the MILP correctly encodes the location objective without distorting the mean-risk trade-off and if the experiments demonstrate advantages over sequential methods, the unified model would provide a computationally attractive way to incorporate network clustering directly into portfolio decisions, extending the literature on correlation-based asset networks.

major comments (2)
  1. [Formulation section (around the definition of the combined objective)] The central modeling claim rests on the assertion that adding a location-problem objective 'measures the effect of clustering on the selected assets with respect to the non-selected ones'; however, the manuscript provides no derivation or numerical check showing that this term remains meaningful once the mean-risk objectives are active (e.g., it may be dominated or become redundant).
  2. [Computational experiments section] The abstract states that 'preliminary computational experiments on real financial datasets validate the approach,' yet the reported results contain no comparison against a two-phase baseline, no out-of-sample performance metrics, and no statistical tests; without these the effectiveness claim cannot be assessed.
minor comments (2)
  1. [Network construction paragraph] Notation for the correlation-derived distance should be introduced once and used consistently; currently the transition from correlation matrix to network metric is described only informally.
  2. [Introduction and model section] The manuscript should cite the specific location problems (p-median, p-center, etc.) whose objective functions are being adapted, rather than referring generically to 'classical location problems on networks.'

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and indicate the revisions we will make.

read point-by-point responses

  1. Referee: [Formulation section (around the definition of the combined objective)] The central modeling claim rests on the assertion that adding a location-problem objective 'measures the effect of clustering on the selected assets with respect to the non-selected ones'; however, the manuscript provides no derivation or numerical check showing that this term remains meaningful once the mean-risk objectives are active (e.g., it may be dominated or become redundant).

    Authors: We agree that the current manuscript does not include a formal derivation or numerical verification of the location term's contribution when the mean-risk objectives are simultaneously active. In the revised version we will add an analysis of the combined objective, including a decomposition showing the scaling and interaction of terms, together with small-scale numerical examples that isolate the effect of the location component. revision: yes

  2. Referee: [Computational experiments section] The abstract states that 'preliminary computational experiments on real financial datasets validate the approach,' yet the reported results contain no comparison against a two-phase baseline, no out-of-sample performance metrics, and no statistical tests; without these the effectiveness claim cannot be assessed.

    Authors: The experiments reported are preliminary and focus on demonstrating computational feasibility of the unified MILP. We acknowledge the absence of direct two-phase baselines, out-of-sample metrics, and statistical tests. The revised manuscript will expand this section to include comparisons against a standard cluster-then-select procedure, out-of-sample evaluation, and statistical significance tests on the performance differences. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper's core contribution is the construction of a new MILP model that augments a standard mean-risk portfolio objective with an additional location-problem term on a correlation network; this is presented as a modeling unification solved in one phase rather than two. No derivation chain, parameter fitting, or uniqueness claim is described that reduces by construction to prior fitted values or self-citations. The approach is self-contained as an optimization formulation whose validity rests on the explicit MILP constraints and objective, not on any recursive definition or imported ansatz.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; the correlation metric and location objective are treated as standard inputs from prior literature.

pith-pipeline@v0.9.0 · 5774 in / 1008 out tokens · 21763 ms · 2026-05-24T21:20:33.006621+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

45 extracted references · 45 canonical work pages

  1. [1]

    M., Heath, D.: Coheren t Measures of Risk

    Artzner, P ., Delbaen, F., Eber, J. M., Heath, D.: Coheren t Measures of Risk. Mathematical Finance 9, 203–228 (1999)

    work page 1999

  2. [2]

    The European Journal of Finance 23, 487–506 (2017)

    Bellini, F., Di Bernardino, E.: Risk management with exp ectiles. The European Journal of Finance 23, 487–506 (2017)

    work page 2017

  3. [3]

    Computers & Operations Research 35, 766–775 (2008)

    Benati, S.: Categorical data fuzzy clustering: an analy sis of local search heuristics. Computers & Operations Research 35, 766–775 (2008)

    work page 2008

  4. [4]

    Comput- ers & Operations Research 43, 280–285 (2014)

    Benati, S., García, S.: A mixed integer linear model for c lustering with variable selection. Comput- ers & Operations Research 43, 280–285 (2014)

    work page 2014

  5. [5]

    Journal of the Operationa l Research Society 69, 1379–1395 (2018)

    Benati, S., García, S., Puerto, J.: Mixed integer linear programming and heuristic methods for feature selection in clustering. Journal of the Operationa l Research Society 69, 1379–1395 (2018)

    work page 2018

  6. [6]

    T OP 22, 934–949 (2014)

    Beraldi, P ., Bruni, M.E.: A clustering approach for scen ario tree reduction: an application to a stochastic programming portfolio optimization problem. T OP 22, 934–949 (2014)

    work page 2014

  7. [7]

    Annals of Operations Research 216, 23–34 (2014)

    Boginski, V ., Butenko, S., Shirokikh, O., Trukhanov, S., Lafuente, J.G.: A network-based data min- ing approach to portfolio selection via weighted clique rel axations. Annals of Operations Research 216, 23–34 (2014)

    work page 2014

  8. [8]

    European Journal of Op erational Research 259, 322–329 (2017)

    Bruni, R., Cesarone, F., Scozzari, A., Tardella, A.: On e xact and approximate stochastic dominance strategies for portfolio selection. European Journal of Op erational Research 259, 322–329 (2017)

    work page 2017

  9. [9]

    D ata in Brief 8, 858–862 (2016)

    Bruni, R., Cesarone, F., Scozzari, A., Tardella, A.: Rea l-world datasets for portfolio selection and solutions of some stochastic dominance portfolio models. D ata in Brief 8, 858–862 (2016)

    work page 2016

  10. [10]

    Annals of Operations Resear ch 205, 213–234 (2013)

    Cesarone, F., Scozzari, A., Tardella, A.: A new method f or mean-variance portfolio optimization with cardinality constraints. Annals of Operations Resear ch 205, 213–234 (2013)

    work page 2013

  11. [11]

    Journal of Global Optimization 68, 439–461 (2017)

    Cesarone, F., Tardella, A.: Equal Risk Bounding is bett er than Risk Parity for portfolio selection. Journal of Global Optimization 68, 439–461 (2017)

    work page 2017

  12. [12]

    qu adratic portfolio selection models with hard real world constraints

    Cesarone, F., Scozzari, A., Tardella, A.: Linear vs. qu adratic portfolio selection models with hard real world constraints. Computational Management Science 12, 345–370 (2015)

    work page 2015

  13. [13]

    Computers & Operations Research 27, 1271–1302 (2000)

    Chang, T.J., Meade, N., Beasley, J.E., Sharaiha, Y .M.: Heuristics for cardinality constrained port- folio optimisation. Computers & Operations Research 27, 1271–1302 (2000)

    work page 2000

  14. [14]

    K., Ziemba, W

    Chopra, V . K., Ziemba, W. T.: The effect of errors in mean s, variances, and covariances on optimal portfolio choice. In Handbook of the Fundamentals of Financ ial Decision Making: Part I, pp. 365- 373 (2013)

    work page 2013

  15. [15]

    P ., Grassi, R., Hitaj, A.: Asset allocatio n: new evidence through network approaches

    Clemente, G. P ., Grassi, R., Hitaj, A.: Asset allocatio n: new evidence through network approaches. Annals of Operations Research. https://doi.org/10.1007/ s10479-019-03136-y (2019)

    work page 2019

  16. [16]

    Discrete Applied Mathematics 9, 27–39 (1984)

    Corneil, D.G., Perl, Y .: Clustering and domination in p erfect graphs. Discrete Applied Mathematics 9, 27–39 (1984)

    work page 1984

  17. [17]

    Discrete Mathematics 86, 145–164 (1990) 21

    Corneil, D.G., Stewart, L.K.: Dominating sets in perfe ct graphs. Discrete Mathematics 86, 145–164 (1990) 21

    work page 1990

  18. [18]

    DeMiguel, V ., Garlappi, L., Uppal, R.: Optimal versus n aive diversification: How inefficient is the 1/n portfolio strategy? The Review of Financial studies 22, 191 5–1953 (2007)

    work page 1953

  19. [19]

    European Financial Management 12, 3 03–318 (2006)

    Elton, E.J., Gruber, M.J., Spitzer, J.: Improved estim ates of correlation coefficients and their impact on optimum portfolios. European Financial Management 12, 3 03–318 (2006)

    work page 2006

  20. [20]

    Ehrgott, M.: Multicriteria Optimization. V ol. 491. Sp ringer Science & Business Media, Berlin Heidelberg (2005)

    work page 2005

  21. [21]

    Annals of Operations Research 147, 343–360 (2006)

    Ehrgott, M.: A discussion of scalarization techniques for multiple objective integer programming. Annals of Operations Research 147, 343–360 (2006)

    work page 2006

  22. [22]

    INFORMS Journal on Computing 23, 546–556 (2011)

    García, S., Labbé, M., Marín, A.: Solving large p-median problems with a radius formulation. INFORMS Journal on Computing 23, 546–556 (2011)

    work page 2011

  23. [23]

    Garey, M.R., Johnson, D.S.: Computers and intractabil ity: A guide to the theory of NP- Completeness. W.H. Freeman and Company, New Y ork (1979)

    work page 1979

  24. [24]

    : Solving large p-median clustering prob- lems by primal-dual variable neighborhood search

    Hansen, P ., Brimberg, J., Urosevic, D., Mladenovic, N. : Solving large p-median clustering prob- lems by primal-dual variable neighborhood search. Data Min ing and Knowledge Discovery 19, 351–375 (2009)

    work page 2009

  25. [25]

    The Journal of Finance 56, 699–20 (2001)

    Jegadeesh, N., Titman, S.: Profitability of momentum st rategies: An evaluation of alternative ex- planations. The Journal of Finance 56, 699–20 (2001)

    work page 2001

  26. [26]

    D., Korkie, B.: Estimation for Markowitz effi cient portfolios

    Jobson, J. D., Korkie, B.: Estimation for Markowitz effi cient portfolios. Journal of the American Statistical Association 75, 544-554 (1980)

    work page 1980

  27. [27]

    II: The p- Medians

    Kariv, O., Hakimi, L.: An Algorithmic Approach to Netwo rk Location Problems. II: The p- Medians. SIAM Journal on Applied Mathematics 37, 539–560 (1 979)

    work page

  28. [28]

    Management science, 37, 519–53 1 (1991)

    Konno, H., Y amazaki, H.: Mean-absolute deviation port folio optimization model and its applica- tions to Tokyo stock market. Management science, 37, 519–53 1 (1991)

    work page 1991

  29. [29]

    Physical Review E 6 6, 026125 (2002)

    Kullmann, L., Kertesz, J., Kaski, K.: Time-dependent c ross-correlations between different stock returns: A directed network of influence. Physical Review E 6 6, 026125 (2002)

    work page 2002

  30. [30]

    Physica A: Statist ical Mechanics and its Applications 287, 412–419 (2000)

    Kullmann, L., Kertesz J., Mantegna, R.: Identification of clusters of companies in stock indices via Potts super-paramagnetic transitions. Physica A: Statist ical Mechanics and its Applications 287, 412–419 (2000)

    work page 2000

  31. [31]

    The Journal of Portfolio Management 36, 60–70

    Maillard, S., Roncalli, T., Telietche, J.: The propert ies of equally weighted risk contribution port- folios. The Journal of Portfolio Management 36, 60–70

    work page

  32. [32]

    G.: Conditiona l value at risk and related linear program- ming models for portfolio optimization

    Mansini, R., Ogryczak, W., Speranza, M. G.: Conditiona l value at risk and related linear program- ming models for portfolio optimization. Annals of operatio ns research 152, 227–256 (2007)

    work page 2007

  33. [33]

    Mansini, R., Ogryczak, W., Speranza, M.G.: Linear and M ixed Integer Programming for Portfolio Optimization, Springer, Switzerland (2015)

    work page 2015

  34. [34]

    The European Physical Journal B- Condensed Matter and Complex Systems 11, 193–197 (1999)

    Mantegna, R.N.: Hierarchical structure in financial ma rkets. The European Physical Journal B- Condensed Matter and Complex Systems 11, 193–197 (1999)

    work page 1999

  35. [35]

    The Journal of Fin ance 7, 77–91 (1952) 22

    Markowitz, H.: Portfolio selection. The Journal of Fin ance 7, 77–91 (1952) 22

    work page 1952

  36. [36]

    Journal of Financial Economics 8, 323–361 (1980)

    Merton, R.C.: On estimating the expected return on the m arket: An exploratory investigation. Journal of Financial Economics 8, 323–361 (1980)

    work page 1980

  37. [37]

    Mirchandani, P .B., Francis, R.L.: Discrete location t heory, Wiley-Interscience Series in Discrte Mathematics and Optimization, New Y ork (1990)

    work page 1990

  38. [38]

    Ng, R.T., Han, J.: Efficient and effective clustering me thods for spatial data mining, in Proceedings of the 20th International Conference on V ery Large Data Base s, VLDB ’94, San Francisco, CA, USA, 1994, Morgan Kaufmann Publishers Inc., pp. 144-155

    work page 1994

  39. [39]

    Onnela, J

    J.-P . Onnela, J. P ., Kaski, K., Kertesz J.: Clustering a nd information in correlation based financial networks. The European Physical Journal B 38, 353–362 (2004 )

    work page 2004

  40. [40]

    T., Uryasev, S.: Conditional value-at -risk for general loss distributions

    Rockafellar, R. T., Uryasev, S.: Conditional value-at -risk for general loss distributions. Journal of banking & finance 26, 1443–1471 (2002)

    work page 2002

  41. [41]

    Journal of Busi ness 39, 119–138 (1966)

    Sharpe, W.F.: Mutual fund performance. Journal of Busi ness 39, 119–138 (1966)

    work page 1966

  42. [42]

    The Journal of Portfoli o Management 21, 49–58 (1996)

    Sharpe, W.F.: The sharpe ratio. The Journal of Portfoli o Management 21, 49–58 (1996)

    work page 1996

  43. [43]

    Management Science 43, 1437–144 6 (1997)

    Simaan, Y .: Estimation risk in portfolio selection: Th e Mean V ariance Model and the Mean- Absolute Deviation model. Management Science 43, 1437–144 6 (1997)

    work page 1997

  44. [44]

    Jour- nal of Economic Dynamics and Control 32, 235–258 (2008)

    Tola, V ., Lillo, F., Gallegati, M., Mantegna, R.N.: Clu ster analysis for portfolio optimization. Jour- nal of Economic Dynamics and Control 32, 235–258 (2008)

    work page 2008

  45. [45]

    Nature 393(6684), 440 (1998) 23

    Watts, D.J., Strogatz, S.H.: Collective dynamics of sm all-worldnetworks. Nature 393(6684), 440 (1998) 23

    work page 1998