Method, system and program product for determining objective function coefficients of a mathematical programming model

A process for deploying computing infrastructure, comprising integrating computer-readable code into a computing system, wherein the code in combination with the computing system is capable of performing a method of determining a plurality of coefficients of an objective function of a mathematical programming model in a computing environment, said method comprising:

• identifying a plurality of attributes of said model, wherein said plurality of attributes contributes to a goal, wherein said goal is a determination of said plurality of coefficients of said objective function, wherein said goal is included in a top level of a hierarchy having a structure specified by the Analytic Hierarchy Process (AHP), wherein said plurality of attributes is included in a second level of said hierarchy, and wherein said second level is subordinate to said top level in said hierarchy;
• determining a first set of objective function coefficient values as initially representing said plurality of coefficients of said objective function, wherein said first set of objective function coefficient values specifies a first solution of said model;
• initializing a prevailing solution to said first solution, wherein said first set of objective function coefficient values determines said prevailing solution;
• generating multiple sets of objective function coefficient values determining corresponding multiple solutions of said model in addition to said prevailing solution, wherein said generating multiple sets of objective function coefficient values includes sampling a sample space based on said prevailing solution, wherein said multiple solutions are included in a bottom level of said hierarchy, wherein said bottom level is subordinate to said second level, and wherein said hierarchy relates each attribute of said plurality of attributes to said multiple solutions;
• executing the AHP by a computing system, wherein said executing the AHP includes evaluating said multiple solutions to provide a ranking of said multiple solutions, wherein said ranking of said multiple solutions is dependent upon a pair-wise comparison of said plurality of attributes across said second level of said hierarchy and a plurality of pair-wise comparisons of said multiple solutions across said bottom level of said hierarchy;
• selecting a second solution of said multiple solutions, wherein said selecting is based on said ranking of said multiple solutions, wherein said second solution is specified by a second set of objective function coefficients;
• determining that said second solution exceeds a sum of said prevailing solution and a specified tolerance;
• setting said prevailing solution to said second solution in response to said determining that said second solution exceeds said sum;
• setting said first set of objective function coefficient values to said second set of objective function coefficient values in response to said determining that said second solution exceeds said sum;
• subsequent to said determining that said second solution exceeds said sum, said setting said prevailing solution, and said setting said first set of objective function coefficient values, repeating said generating, said executing the AHP, said selecting said second solution, said setting said prevailing solution, and said setting said first set of objective function coefficient values until said second solution does not exceed said sum;
• determining, subsequent to said repeating, that said plurality of coefficients of said objective function is said first set of objective function coefficient values;
• and
• storing, by said computing system and subsequent to said repeating, said first set of objective function coefficient values in a data storage device of said computing environment.

A computer-implemented method of determining a plurality of coefficients of an objective function of a mathematical programming model, said method comprising:

• receiving, by a computing system, a goal, a plurality of decision-making criteria, and a prevailing solution of said mathematical programming model, wherein said goal is a determination of said plurality of coefficients of said objective function, wherein said plurality of decision-making criteria contributes to said goal, wherein said goal and said plurality of decision-making criteria are required by the Analytic Hierarchy Process (AHP), and wherein said prevailing solution is specified by a first set of objective function coefficient values;
• generating, by said computing system, multiple sets of objective function coefficient values that specify, in a one-to-one correspondence, multiple solutions of said mathematical programming model, wherein said multiple solutions are based on said prevailing solution, and wherein said plurality of decision-making criteria relate each solution of said multiple solutions to said goal according to the AHP;
• generating an AHP hierarchy as a hierarchical structure specified by the AHP, wherein said AHP hierarchy includes at least three levels, wherein said three levels includes a top level, a second level subordinate to said top level, and bottom level subordinate to said second level, wherein said top level includes said goal, wherein said second level includes said plurality of decision-making criteria, wherein said bottom level includes said multiple solutions, and wherein said AHP hierarchy relates each decision-making criterion of said plurality of decision-making criteria to said multiple solutions;
• executing the AHP by said computing system, wherein said executing the AHP includes evaluating said multiple solutions, wherein said evaluating includes:
• pair-wise comparing, across said second level of said AHP hierarchy, said decision-making criteria included in said plurality of decision-making criteria;
• determining, in response to said pair-wise comparing said decision-making criteria, a first plurality of weights, wherein a weight of said first plurality of weights indicates, in relation to achieving said goal, a degree of importance of one decision-making criterion of said plurality of decision-making criteria over another decision-making criterion of said plurality of decision-making criteria, and wherein each weight of said first plurality of weights is independent of a value resulting from said objective function;
• pair-wise comparing said multiple solutions across said bottom level of said AHP hierarchy for each decision-making criterion of said plurality of decision-making criteria;
• determining, in response to said pair-wise comparing said multiple solutions, a second plurality of weights, wherein a weight of said second plurality of weights indicates, in relation to satisfying a decision-making criterion of said plurality of decision-making criteria, a degree of importance of one solution of said multiple solutions over another solution of said multiple solutions, and wherein each weight of said second plurality of weights is independent of said value resulting from said objective function;
• determining, subsequent to said determining said first plurality of weights and said determining said second plurality of weights, a plurality of aggregate values based on an execution of the AHP operating on said first plurality of weights and said second plurality of weights,
• determining, subsequent to said determining said plurality of aggregate values, multiple rankings of said multiple solutions based on said aggregate values;
• and
• selecting, subsequent to said determining said multiple rankings, a second solution of said mathematical programming model, wherein said second solution is included in said multiple solutions, wherein said second solution is specified by a second set of objective function coefficient values, and wherein said second solution has a ranking of said multiple rankings that indicates a superiority of said second solution over any other solution of said multiple solutions;
• determining, by said computing system, that said second solution exceeds a sum of said prevailing solution and a specified tolerance;
• setting, by said computing system and in response to said determining that said second solution exceeds said sum, said prevailing solution to said second solution;
• setting, by said computing system and in response to said determining that said second solution exceeds said sum, said first set of objective function coefficient values to said second set of objective function coefficient values;
• repeating, by said computing system, said generating multiple sets of objective function coefficient values, said evaluating, said setting said prevailing solution, and said setting said first set of objective function coefficient values until said second solution does not exceed said sum;
• determining, by said computing system and subsequent to said repeating, that said plurality of coefficients of said objective function is said first set of objective function coefficient values;
• and
• storing, by said computing system and subsequent to said repeating, said first set of objective function coefficient values in a data storage device.

A computer-implemented method of determining objective function coefficient values of a linear programming (LP) model, said method comprising:

• generating, by a computing system, a first set of tentative objective function coefficient values of a plurality of objective function coefficients of said LP model, a tentative optimal solution of said LP model, and a set of linear inequalities over a space of said plurality of objective function coefficients, wherein said tentative optimal solution is based on said first set of tentative objective function coefficient values;
• performing, via a first execution of the Simplex Algorithm by said computing system and subsequent to said generating said tentative optimal solution, a first pivot, wherein said performing said first pivot includes pivoting a basis of said tentative optimal solution, and wherein a result of said performing said first pivot is a first alternate feasible solution of said LP model;
• determining, by said computing system and subsequent to said performing said first pivot, a first difference by subtracting an objective function value of said first alternate feasible solution from an objective function value of said tentative optimal solution;
• determining, by said computing system and subsequent to said determining said first difference, that a first linear inequality representing a non-negativity of said first difference is not a redundant linear inequality with respect to said set of linear inequalities;
• receiving, by said computing system and subsequent to said determining that said first linear inequality is not said redundant linear inequality, a first preference between said tentative optimal solution and said first alternate feasible solution;
• inserting, by said computing system and subsequent to said receiving said first preference, a second linear inequality into said set of linear inequalities, wherein said second linear inequality represents said first preference;
• determining, by said computing system and subsequent to said inserting said second linear inequality, that said first preference indicates a preference of said tentative optimal solution over said first alternate feasible solution;
• repeating said performing said first pivot, said determining said first difference, said determining that said first linear inequality is not said redundant linear inequality, said receiving said first preference, and said inserting said second linear inequality until said first preference indicates a preference of said first alternate feasible solution over said tentative optimal solution;
• determining, by said computing system and subsequent to said repeating, that said first preference indicates a preference of said first alternate feasible solution over said tentative optimal solution;
• updating, by said computing system and in response to said determining that said first preference indicates said preference of said first alternate feasible solution over said tentative optimal solution, said first set of tentative objective function coefficient values to generate a second set of tentative objective function coefficient values and to make said first alternate feasible solution optimal, wherein said second set of tentative objective function coefficient values maintains a consistency with said set of linear inequalities;
• setting, by said computing system and in response to said determining that said first preference indicates said preference of said first alternate feasible solution over said tentative optimal solution, said tentative optimal solution as said first alternate feasible solution;
• performing, via a second execution of the Simplex Algorithm by said computing system and subsequent to said setting said tentative optimal solution as said first alternate feasible solution, a second pivot, wherein said performing said second pivot includes pivoting a basis of said tentative optimal solution, and wherein a result of said performing said second pivot is a second alternate feasible solution of said LP model;
• determining, subsequent to said performing said second pivot, based on an analysis of said set of linear inequalities and not based on a user-determined preference between said second alternate feasible solution and said tentative optimal solution, that said second alternate feasible solution is necessarily not superior to said tentative optimal solution;
• and
• determining, by said computing system and subsequent to said performing said second pivot, that no other pivots of a plurality of pivots are unperformed, wherein said plurality of pivots is associated with said tentative optimal solution;
• identifying, by said computing system and subsequent to said determining that no other pivots of said plurality of pivots are unperformed, said second set of tentative objective function coefficient values as a final set of objective function coefficient values of a final optimal solution of said LP model;
• and
• storing, by said computing system and subsequent to said identifying, said final set of objective function coefficient values in a data storage device coupled to said computing system.