Mean Squared Variance Portfolio: A Mixed-Integer Linear Programming Formulation

The mean-variance (MV) portfolio is typically formulated as a quadratic programming (QP) problem that linearly combines the conflicting objectives of minimizing risk and maximizing expected return through aversion profile parameter. In this formulation, two are expressed in different units, an issue could definitely hamper obtaining more competitive set weights. For example, modification scale which returns (by one or percent) MV portfolio, implies solution problem. Motivated by issue, novel mean squared variance (MSV) proposed paper. associated optimization strategy very similar to Markowitz optimization, with exception mean, presented form our formulation. resulting model non-convex QP problem, has been reformulated mixed-integer linear (MILP) reformulation initial into MILP allows for future researchers practitioners obtain global via use current state-of-the-art solvers. Additionally, purely data-driven method determining optimal value hyper-parameter MSV approaches also empirically tested on eight time series problems three estimation windows (composing total 24 datasets), showing performance most problems.