Decision Making An Introduction 1 Decision Making Decision

Decision Making: An Introduction 1

Decision Making • Decision Making is a process of choosing among two or more alternative courses of action for the purpose of attaining a goal or goals. • It is influenced by several major disciplines which are behavioral and scientific in nature. • Behavioral disciplines include anthropology, law, philosophy, political science, psychology, social psychology, and sociology. • Scientific disciplines include computer science, decision analysis, economics, engineering, management science/operations research, mathematics and statistics. 2

Logical flow of a decision making/problem solving process Criteria Environment Alternatives Problem Decision • Alternatives: possible actions aimed at solving the given problem • Criteria: measurements of effectiveness of the various alternatives and correspond to system performances such as o o o o Profitability Overall cost Productivity Quality Dependability Risk Service Flexibility 3

Logical structure of a decision process Constraints Operational Technical Procedural Legal Social Political Ruled-out decisions Exclusion Alternative options Feasible options Evaluation Criteria Profitability Overall cost Quality Dependability Flexibility Service Decision 4

Models • A main characteristic of a Decision Support System is the inclusion of at least one model. • Model is a selective abstraction of a real system designed to analyze and understand from an abstract point of view the operating behavior of a real system • Includes only elements deemed relevant for the purpose of the investigation being carried out. Real world systems Systems idealized by assumptions Model 5

Different types of models according to their characteristics • Iconic: material representation of a real system, whose behavior is imitated for the purpose of the analysis. Example: a miniaturized model of a new city neighborhood. 6

Different types of models according to their characteristics • Analogical: imitates real behavior by analogy rather than by replication. Example: a wind tunnel built to investigate the aerodynamic properties of a motor vehicle. 7

Different types of models according to their characteristics • Symbolic: abstract representation of a real system. It describes the behavior of the system through a series of symbolic variables, numerical parameters and mathematical relationship. 8

Different types of models according to their probabilistic nature q Stochastic: some input information represents random events, characterized by probability distribution. q Deterministic: all input data are supposed to be known a priori and with certainty. o When it is not possible to know the data with absolute certainty, sensitivity and scenario analyses allow one to test the robustness of the decisions to variations in the input parameters. 9

Different types of models according to their temporal dimension q Static: considers a given system and related decision-making process within a single temporal stage. q Dynamic: considers a given systems through several temporal stages, corresponding to a sequence of decisions. o Discrete-time dynamic models observe the status of a system only at the start or the end of discrete intervals. o Continuous-time dynamic models consider a continuous sequence of periods on the time axis. 10

Development of a model Problem identification Model formulation Feedback Development of algorithms Implementation and testing 11

Development of a model Problem identification Model formulation Observed critical symptoms must be analyzed and interpreted to formulate hypotheses for investigation. Define a mathematical model to represent the system. Important factors: 1. Time horizon Feedback 2. Evaluation criteria: o monetary costs and payoffs Development of algorithms o effectiveness and level of service o quality of products and services o reliability in achieving objectives Implementation and testing o flexibility of operating conditions 3. Decision variables, eg. production volumes. 4. Numerical parameters, eg. production capacity 5. Mathematical relationship 12

Development of a model Problem identification o A solution algorithm is identified o A software tool that incorporates the solution Modelor acquired method should be developed formulation o Analyst should have thorough knowledge of current solution methods and their characteristics o Assess correctness of data and Feedback parameters Development of algorithms o Model validation by experts: • plausibility and likelihood of the Implementation and testing conclusions achieved • consistency of results at extreme values of parameters • stability of results with minor changes in the parameters 13

Classes of models � Predictive models: input data used to predict future events/outcomes. o Regression: a set of independent variables used to predict a continuous dependent variable value, e. g. salary o Classification: a set of independent variables used to predict a discrete dependent variable value, e. g. approve/not approve. � Pattern recognition and machine learning models: efficient algorithms that learn from past observations and derive new rules for the future. o Interpretation models: identify regular patterns, express them as understandable rules and criteria. o Prediction models: forecast future value. o Supervised learning: target is known, e. g. classification, regression. o Unsupervised learning: target does not exist, e. g. clustering. 14

Classes of models �Optimization models: given a set of feasible decisions, identify the optimal one according to the chosen evaluation criterion �Different forms of optimization models: o Linear optimization o Integer optimization o Convex optimization o Network optimization o Multiple-objective optimization 15

Classes of models � Project management models: a set of interrelated activities carried out in pursuit of a specific goal, e. g. a new product line. Network models are often used to represent the component activities of a project and the precedence relationships among them. � Risk analysis models: choose among a number of available alternatives, having uncertain information regarding the effects that these options may have in the future. � Waiting line models: to investigate congestion phenomena occurring when the demand for and the provision of a service are stochastic in nature. 16

In this module …. � Optimization models � Predictive models � Pattern recognition and machine learning models 17

References. � Business Intelligence, C. Vercellis, Wiley, 2009. Chapters 2 and 4. � Decision Support and Business Intelligence Systems, Pearson International, 8 th Ed. Chapter 2. 18