Decision modeling: is a scientific approach informal methods
to decision making. 2- The problem is very important It can also be defined as the 3- The problem is new development of a model (usually 4- The problem is repetitive mathematical) of a real-world problem Because you do not need to start from scenario or environment. the beginning each time the similar The scientific approach is: problem occurs, since you can use the 1- Without biases and emotions model created by quantitative analysis 2- Tested approach. 3- Systematic Decision models: Decision modeling is also commonly 1- Are Generally mathematical referred to as quantitative analysis, representations. management science, operations 2- Provide analytical framework for research, or operational research. evaluating modern business The correct use of decision modeling problems. techniques usually results in solutions 3- Provide techniques applicable in that are many areas 1- Timely Production & operations management, 2- Accurate supply chain management, accounting, 3- Flexible logistics, marketing, finance, etc. 4- Economical Decision models are subject to: 5- Reliable 1- Limitations 6- Easy to understand 2- Assumptions 7- Easy to use Assuming is different than forecasting Management science had its early roots in and predicting. World War II and is flourishing in business Assumption is not evidence-based, it is and industry with the aid of computers. without proof. In WWII, it was used to choose targets 3- Simplifications to shoot because army had limited Using assumptions real-world resources. problems are simplified into solvable Management science provides insights virtual ones. into the solution of the managerial Models are not the real problems; they problem: are just abstractions of them. 1- Whether the solution is beneficial d ifi S ti sn p w a (th m le b ro )fu c D R g or not 2- Whether it is robust or not 3- Whether it is sensitive or not analysis interpretation Management science persons role is being a decision/system analyst. MANAGERIAL Management science person is the JUDGEMENT bridge between the problem and the decision maker. Management science person supports abstraction intuition
and guides the decision maker.
We use quantitative analysis approach to decision making when: Results of the model cannot directly 2- Models could be really hard to implemented to the real world problem. solve if not impossible Variable: a measurable quantity that is 3- Can lose the real problem subject to change. Too much abstraction/assumptions Variables can be controllable or 4- Quantitative analysis in expense of uncontrollable. qualitative analysis Decision variable: a controllable variable A complex model will give more accurate Parameter: a measurable quantity that is solutions but it is more costly and hard to inherent in the problem. solve. Usually has a known value. So, it is NOT APPROPRIATE! Some kind of uncontrollable. A less complicated model will be less In most cases, variables are unknown precise but it is less costly and easier to quantities, whereas parameters (or input solve. data) are known quantities. So, it is APPROPRIATE! Parameter is a data as a number, and Decision models can be broadly classified constraint is the statement that includes into two categories, based on the type it. and nature of the decision-making problem environment under consideration: 1- deterministic models 2- probabilistic models. Relate decision variables (controllable Deterministic: means with complete inputs) with fixed or variable parameters certainty. (uncontrollable inputs). Deterministic models assume: Frequently seek to maximize or 1- All the relevant input data values minimize some performance measures are known with certainty (objective function) subject to constraints. 2- All the information needed for In other words, objective function modeling a decision-making must have a direction. problem environment is available, The values of the decision variables with fixed and known values. that provide the mathematically-best Most commonly used deterministic output are referred to as the optimal modeling technique is linear solution for the model. programming. Advantages of decision models are: Probabilistic (stochastic) models assume: 1- No emotion/no bias 1- Some input data values are not 2- Consistent known with certainty 3- A systematic approach 2- The values of some important 4- Easy to express/easy to deal with variables will not be known before 5- Easy to experiment on decisions are made Experimenting with models when it It is important to incorporate this compared to experimenting with the real “ignorance” into the model. situation are: Examples of probabilistic models are; the 1- Require less time decision of whether to start a new 2- Less expensive business venture, and the decision of 3- Involve less risk which major to choose when students Disadvantages of decision models are: enter the college. 1- Constructing models could be hard Probabilistic modeling techniques provide the required input data should be a structured approach for managers to obtainable. incorporate uncertainty into their models The model should include the and to evaluate decisions under alternate appropriate amount of detail for the expectations regarding this uncertainty. model to be solvable yet realistic. The decision modeling process involves Usually, it is a mathematical model. three distinct steps: 3- Acquiring input data 1- Formulation Acquired data should be accurate. 2- Solution Company reports and documents can 3- interpretation. be used to obtain the necessary data. It is common to have an iterative Sampling and direct measurement can process between these three steps before be used as a source of data. the final solution is obtained. Statistical sampling procedures can be used to obtain data. Managers can be used as a source of data. Solution: is when the mathematical expressions resulting from the formulation process are actually solved to identify the optimal solution. Solution can be classified into two parts: 1- Developing a solution Involves manipulating the model to arrive at the best (or optimal) solution to the problem. Formulation: is the process by which each May require that a set of mathematical aspect of a problem scenario is translated expressions be solved to determine the and expressed in terms of a mathematical best decision. model. Trial and error method can be used Formulation is the most challenging step In some cases all possible values for in decision modeling. the variables should be tried to arrive the Formulation can be classified into three best solution (complete enumeration) parts: Complex and difficult problems need 1- Defining the problem the utilization of an algorithm/heuristic Develop a clear, concise statement of The accuracy of the solution depends the problem. on the accuracy of the input data and the The problem statement gives direction decision model itself. and meaning to all the parts that follow it. 2- Testing the solution It is essential to go beyond just the The input and the data model both symptoms of the problem at hand and require testing. identify the true causes behind it. Input data can be tested by collecting When a problem is difficult to quantify, additional data from a different source it may be necessary to develop specific, and using statistical tests to compare measurable objectives. these new data with the original data. 2- Developing a model The model should be checked to make The model should be solvable, realistic, sure that it is logical and represents the and easy to understand and modify, and real situation. Algorithm: is a series of steps that are to the wrong problem, so beginning repeated until attaining a satisfactory assumptions should be carefully assumed. solution. 5- Problems that change quickly Interpretation can be classified into two Even if a problem has been specified parts: correctly at present, it can change during 1- Analyzing the results and the development of the model. sensitivity analysis Possible problems that can be faced Since the model is only an during developing the model: approximation of reality, the sensitivity of 1- Fitting textbook models the solution to changes in the model and A manager’s perception of a problem input data should be analyzed by does not always match the textbook conducting a sensitivity (what-if) analysis. approach. When the optimal solution is very 2- Understanding/accepting a model sensitive to changes in the input data and Managers do not use the results of a the model specifications, additional model they do not understand, so the testing must be performed to make sure model should be easier to understand. the model and input data are accurate Possible problems that can be faced and valid. during acquiring data: 2- Implementing the results 1- Availability of the data optimal solution should not 2- Accessibility to the data implemented directly, it is only to give 3- Relevance of the data insight. 4- Quality of the data Solution implementation usually Data should be good and clean requires making changes within the 5- Missing data organization. Possible problems that can be faced Managers of the company can resist to during developing the solution: the changes that should made. 1- Hard-to understand/complex The solution should be closely mathematics monitored even after implementation. Although the mathematical models we Sensitivity analysis: determines how the use may be complex and powerful, they solutions will change with a different may not be completely understood. model or input data. 2- The limitation of giving only one Possible problems that can be faced answer during defining the problem: Decision models usually give just one 1- Conflicting viewpoints answer to a problem, so insight is The problem needs to be examined important. from several viewpoints. Possible problems that can be faced 2- Impact on other departments during testing a solution: The problem statement should be as 1- Complex models tend to give broad as possible and include inputs from solutions that are not intuitively all concerned departments. obvious. 3- Real life is too ambiguous– a These type of solutions tend to be jungle rejected by managers, so decision analyst 4- Beginning assumptions should review the assumptions and A “good” solution to the right problem convince the manager of the validity of is much better than an “optimal” solution the solutions. Possible problems that can be faced The process by which each aspect of a during analyzing the results: problem scenario is translated and 1- The results must be analyzed in expressed in terms of simple terms of how they will affect the mathematical expressions. total organization. 2- Solution The analyst should ascertain who must The mathematical expressions change and by how much, if the people resulting from the formulation process who must change will be better or worse are solved to identify an optimal (or best) off, and who has the power to direct the solution to the model. change. 3- Interpretation Possible problems that can be faced Allow the manager to evaluate the during implementing the results: impact of several different types of what- 1- Resistance for change if questions regarding the problem. If results suggest large changes in Properties of an LP model: organizational policy, the decision analyst 1- Objective function can expect resistance. Problems seek to maximize or Management support and user minimize an objective which is usually involvement are important in minimizing cost or maximizing profit. implementing results 2- Constraints 2- Time dimension limit the degree to which we can There can be no immediate effects. pursue our objective. CHAPTER 2 We want to maximize or minimize a The most widely used decision modeling quantity (the objective) subject to limited technique designed to help managers in resources. resource allocation is called mathematical An LP model usually includes a set of programming. constraints known as nonnegativity The most widely used modeling constraints. technique designed to help managers in 3- Alternatives planning and decision making is linear There must be alternative courses of programming (LP). action from which we can choose. When developing LP (and other 4- Linearity mathematical programming)–based The objective and constraints in LP decision models, we assume that all the problems must be expressed in terms of relevant input data and parameters are linear equations or inequalities. known with certainty. In linear mathematical relationships, all LP has been applied extensively to: terms used in the objective function and 1- Medical problems constraints are of the first degree. 2- Transportation problems Feasible solution: satisfies all the 3- Operations problems problem's constraints. 4- Financial problems Optimal solution: satisfies all the 5- Marketing problems problems constraints, and results in the 6- Accounting problems largest possible objective function values 7- Human resources problems when maximizing / results in the 8- Agricultural problems. minimum possible objective value when in different sizes and complexities. minimizing. LP models have three steps: All optimal solutions are feasible! 1- Formulation Not all feasible solutions are optimal! Assumptions of an LP model: The objective function states the goal 1- Certainty of a problem / the motivation for solving a Numbers used in the objective function problem. and constraints are known with certainty An LP model must have a single and do not change during the period objective function. being studied. 3- Constraints 2- Proportionality Constraints represent restrictions on Exists in the objective function and the values the decision variables can take. constraints. Each constraint is expressed as a 3- Additivity mathematical expression and can be The total of all activities equals the independent of the other constraints in sum of the individual activities. the model. 4- Divisibility It is very common in practice for a Solutions need not necessarily be in single LP model to include constraints whole numbers (integers). with different signs (i.e. , ≤ , and = ). Solutions may take any fractional 4- A key principle of LP is that value. interactions exist between If a fraction of a product cannot be variables. produced, an integer programming 5- Nonnegativity constraints specify problem exists. that decision variables cannot Problem formulation (modeling): is the have negative values. process of translating a verbal statement 6- In LP models, we do not specify of a problem into a mathematical that decision variables should only statement. have integer values. Steps in problem formulation: A model in which some or all of the 1- Define the decision variables decision variables are restricted only to 2- Construct the objective function integer values is called an integer 3- Formulate the constraints programming (IP) model. One of the most common LP applications IP models do not allow detailed is the product mix problem. sensitivity analysis whereas LP models Product mix problems use LP allow detailed sensitivity analysis. to decide how much of each product to 7- To find an optimal solution to an make, given a series of resource LP problem, we must first identify restrictions. a set, or region, of feasible The structured approach for formulating solutions. LP problems: The first step in doing so is to plot each 1- Decision variables of the problem’s constraints on a graph. Decision variables are the unknown We can plot either decision variable on entities in a problem. The problem is the horizontal (X) axis of the graph, and solved to find values for decision the other variable on the vertical (Y) axis. variables. The nonnegativity constraints imply Decision variables are expressed in the that we are working only in the first (or problems using alphanumeric symbols. positive) quadrant of a graph. Different decision variables in the same The feasible region of minimization model can be measured in different units. problems is often unbounded (i.e., open 2- The objective function outward). Identifying optimal solution by: 1- Drawing level lines No feasible solution region exists. Also called isocost, isoprofit lines. Infeasibility is not related with We use the objective function to point objective function. us toward the optimal solution. Lack of a feasible solution region can We draw a series of parallel level lines occur if constraints conflict with one until we find the one that corresponds to another. the optimal solution. Excel shows infeasibility! An optimal solution to an LP model If we have infeasibility, we cannot must lie at one of the corner points in the solve the problem. feasible region. 3- Alternate optima Solving for the coordinates of a corner There is more than one optimal point requires the use of simultaneous solution. equations. Occurs when a binding constraint (that 2- Corner point method is a constraint defining the optimal eliminates the need for graphing and solution) and the objective function have using level objective function lines same slope. Identifies an optimal solution by Far from causing problems, the evaluating the objective function value at presence of more than one optimal every corner point in the problem. solution actually allows management Notes for developing a correct LP model: greater flexibility in deciding which 1- All terms within the same solution to select. expression must refer to the same The optimal objective function value entity. remains the same at all alternate 2- All terms within the same solutions. expression must be measured in 4- Unboundedness the same units. When a problem has an unbounded 3- Address each constraint feasible region, it may not have a finite separately. solution. There is no single “mega” expression When an LP model has a bounded that will take care of all constraints in the feasible region, it has an identifiable model at one time. optimal corner point solution. Each constraint is a separate issue, and If the feasible region is unbounded in expressions for each must be written one or more directions, depending on the separately. objective function, the model may or may Special situations in LP: not have a finite solution. 1- Redundancy If our objective function is unbounded, A redundant constraint is a constraint our feasible region is also unbounded. that does not affect the feasible solution When the solution is unbounded in a region in any way. maximization problem, the objective If redundant constraint is close to function value can be made infinitely feasible region, it may be important! large without violating any constraints. Excel does not detect redundant When the solution is unbounded in a constraints! minimization problem, the objective 2- Infeasibility function value can be made infinitely A condition that arises when no single minimum without violating any solution satisfies all of an LP problem’s constraints. constraints. In real-world situations, the occurrence In a binding constraint there is no slack or of an unbounded solution usually means surplus. the problem has been formulated CHAPTER 4 improperly. Network flow models are special LP Excel warns unboundedness! models. Applications of LP in businesses: Types of network flow models are: 1- Marketing applications 1- Transportation Media selection problem and 2- Transshipment marketing research problem 3- Assignment 2- Manufacturing applications 4- Maximal-flow Product mix problem 5- Shortest-path 3- Employee scheduling applications 6- Minimal-spanning tree models Labor planning problem Networks consists of nodes and arcs that 4- Financial applications connect nodes together. Portfolio selection problem 5- Transportation applications Truck-loading problem 6- Ingredient blending applications Diet problem and ingredient mix/blending problem 7- Multiperiod applications Product scheduling problem and sinking fund problem In ‘’Answer Report’’ in Excel, ‘’contin’’ means continuous. This situation points out the divisibility characteristic of LP models. In ‘’status’’ in the ‘’Answer Report’’ in Excel: 1- Binding means LHS = RHS It means that I am fully using the constraint. 2- Not binding means LHS is not equal to RHS It means that I am not fully using the constraint. In a less than or equal to (<=) constraint, amount of unused resources called slack. Slack = RHS – LHS In a greater than or equal to (>=) constraint, amount of over satisfaction called surplus/excess. Surplus = LHS – RHS Whether it is a slack or surplus, Excel name both of them as slack in its answer report!