|
|
- Eileen Bryan
- 5 years ago
- Views:
Transcription
1 ENGM 670 & MECE 758 Modeling and Simulation of Engineering Systems (Advanced Topics) Winter 2011 Lecture 10: Extra Material M.G. Lipsett University of Alberta Topics Decision Analysis Uncertainty Model verification Sensitivity Analysis Using Simulations Introduction to Bayes Risk Risk Analysis by simulating the effect of uncertainty on a system with probabilistic variables (Monte Carlo simulation) Probability distribution functions, estimating probabilities and loss functions Probabilistic simulations of discrete-event systems and lumped-parameter systems Model validation (estimating parameters of constitutive relationships in linear and nonlinear systems) MG Lipsett,
2 Why Decisions Are Difficult Many factors affect decisions Technical Issues Financial Issues Management Issues People generally have a poor perception of risk Factors should be weighted to represent their relative importance Not every factor is equally important Appropriate relative weighting is difficult to do MG Lipsett, Decision Analysis Methods Structured methods to determine actions to manage risk Depends on good corporate information on costs, probabilities, and dependencies A facilitation process is used to ensure all scenarios are explored and considered consistently One example is Kepner-Tregoe's thinking processes: Situation Appraisal, Problem Analysis, Decision Analysis, Potential Problem Analysis, and Potential Opportunity Analysis MG Lipsett,
3 Basic Decision Analysis Process Gather experts and decision-makers Describe situation to the group Identify issues relevant to the situation Identify alternatives Usually established in advance (to have good supporting info available) Sometimes brainstorming is done during the process (free development of alternatives) followed by reality check and selection of a small set of realistic alternatives for analysis & decision Good simulations of how an option is expected to perform can be extremely helpful in decision making Select important factors and relative weighting factors For each option or event, estimate value for each factor in terms of costs and costs & benefits MG Lipsett, Example of Weighted Decision Factors Hydro run of river Contaminant Emissions CO 2 Radioactivity Land Use Water Impacts Waste Impacts Resource Availability Total Impact (Weighted) Hydro impoundment Manitoba purchase Wind Biomass Photovoltaic Nuclear Natural gas single cycle Natural gas combined cycle Natural gas cogeneration Gasification without CO 2 removal Gasification with 90% CO 2 removal Coal Oil Note: The Total Weighted Impact is calculated by applying a weight of 10 to contaminant emissions, 20 to greenhouse gases, and 1 to all other categories. Source: SENES and OPA MG Lipsett,
4 Where Decision Analysis Can Go Wrong A decision can go the wrong way for a number of reasons: Bad definition of the problem (trying to make a decision for the wrong problem, or poor definition of what the desired outcome should be and its specifications) Bad assumptions (trying to solve the wrong problem) Poor design of solution options (lack of creativity, constrained thinking) Poor modeling (not producing enough understanding of whether the proposed solution is likely to solve the problem) Poor presentation (communication of understanding is not sufficient to allow decision-makers to assess an option properly) Decision bias (decision-maker has undeclared preference for some options over others, perhaps including not solving the real problem) MG Lipsett, Responsibility for Good Decisions The engineering designer or manager can overcome all of these potential decision-making pitfalls (except the last one) by: careful problem analysis to define the actual need good creative engineering thinking use of model-based analysis for assessing solution options effective packaging of model results for understanding and effective assessment But: A good model, poorly communicated, is worse than useless. MG Lipsett,
5 Uncertainty To this point, we have assumed that we have perfect knowledge about a system. In reality, we may not know everything we need. Uncertainty comes in three forms an observed set of state variables that does not completely define the process state, A constitutive relationship that is not clearly defined, or for which the parameters are not well known, measurements of the system that are inaccurate All three forms occur in design, manufacturing, and other types of operations. Human operators and supervisors must make judgments based on incomplete information and past experience. Uncertainty in a process demands branching for exception handling based on monitoring the task execution. MG Lipsett, Sources of Uncertainty Assumptions about the model Model parameters Measurements of the inputs, outputs, and state variables of the system MG Lipsett,
6 Model Verification Verification is the process of checking that the model is a reasonable approximation of the system of interest A number of techniques can be used MG Lipsett, Model Verification Techniques MG Lipsett, Debug as you develop the simulation Have the simulation reviewed, preferably by more than one person Run the simulation under a number of inputs, and check that the output is reasonable (that is, the model behaves as you would expect) Run a trace to debug a DES; in a trace, the states and certain counters are displayed after each event occurs, allowing hand checks to be made of each step Run the simulation under simplified conditions, a benchmark case with known results, or a simple case with an analytical solution Display an animation of the simulation output Compare the input distribution descriptive statistics (sample mean and sample variance) to historical values, to make sure that the proper inputs are being used in the simulation 6
7 Model Parameters Uncertainty in the model parameters affects the output of the simulation Even if the structure of the governing equations or representations of discrete-events are correct, incorrect parameters drive the simulation to an erroneous set of outputs Parameter estimation can be used to identify the best guess for the parameters for a set of given data (Model validation and nonlinear parameter estimation are discussed later in this slide deck) MG Lipsett, Sensitivity Analysis In some systems there may be some underlying uncertainty in the parameters, especially for a new design In that case, the simulation itself can be used to test how a parameter affects the system behaviour Sensitivity analysis determines how important a variable or a parameter is to the system behaviour Uncertainty can be expressed as a range of values for an aspect of the model (or as a probability) Scenarios are often done (most likely / best /worst case) to give a range of expected results of a change in a value More precise scenario analysis can be done using simulations that test the probabilities using a distribution MG Lipsett,
8 Sensitivity of Individual Parameters For each parameter: Do the analysis first using the most likely value Realistic result Repeat the analysis, assigning the parameter its best-case value Optimistic result showing sensitivity of that parameter Repeat the analysis, assigning the parameter its worst-case value Pessimistic result showing sensitivity of that parameter Results are intuitively visualized using a Tornado Diagram MG Lipsett, Most Likely / Best Case / Worst Case The effect of variability of a parameter (such as a spring constant, transducer coefficient, etc.) can be assessed for its effect on system performance (such as the value of an output of the system) by analysing multiple cases for that alternative Choose the most likely value for the parameter, and a number on either side that represents the best-case value and the worst-case value Ideally these parameters bracket the risk associated with the uncertainty Assess the sensitivity of a parameter individually MG Lipsett,
9 Modeling General Uncertainty Scenarios Do three scenarios (realistic, optimistic, pessimistic) Do the analysis first using the most likely values for all the parameters. This gives the realistic result Repeat the analysis, assigning all the uncertain parameters their best-case values. This gives the most optimistic result Repeat the analysis, assigning all the uncertain parameters their worst-case values. This gives the most pessimistic result Useful for general assessment of the extreme range of possibilities Not very useful for understanding the effect of variability MG Lipsett, MG Lipsett, Risk Associated with An Event Or A Decision The consequence associated with a particular factor can be expressed as a cost (or a benefit, if the outcome is positive) Risk is any uncertainty that affects a system Risk depends on the conditions and time Uncertainties and risks have a time horizon Uncertainties will exist regarding future conditions and states of a system Uncertainties evolve over time Uncertainties become risks when they have consequences, that is, they affect measurable outputs of the system that are set against some standard Risk tends to reduce as time progresses: e.g., marketing uncertainty about how many units of a product goes to zero once all of the units have been sold. 9
10 Conditional Risk & Minimum Risk The conditional risk is defined as: for i = 1,,a, where a is number of possible actions α c is number of states of the system λ(α i w j ) is the loss function for taking an action α i when the state of the system is w j P(w j x) is the posterior probability that the state w j given that the feature vector x has been measured Select the action for which R is minimum (Bayes Risk): MG Lipsett, Expected Outcome for A Set of Probabilities From a set of independent probabilities and associated costs (or benefits), the expected outcome can be estimated easily A utility is planning to build a new electrical generating plant. There are some uncertainties about each technology choice, which will affect the useful life of each. The probabilities are as follows: Probabilities Useful Life Option A Option B Option A will cost $150k per megawatt. Option B will cost $300k per megawatt. Which has the lowest cost $/MWe/yr? MG Lipsett,
11 Probability Distributions How do we represent a range of possibilities? MG Lipsett, Probability Distribution Functions A random variable (RV), X, is a real numerical valued function defined over a sample space, S. X(s) assigns a real number to every outcome in S. A continuous random variable has a range with an infinite number of values on the real line and is typically (but not necessarily) defined over an infinite sample space. The probability density function (pdf) for a continuous random variable is a function that assigns a probability density to each and every value of the random variable. The probability distribution function (pdf) for a discrete random variable is a function that assigns a probability to each value of the random variable. A sampled sequence of a continuous RV is a set of discrete RVs. MG Lipsett,
12 Examples of PDFs Exponential Gaussian Discrete RVs (e.g. sampled data) have discrete PDFs Uniform Beta Gamma PDFs can be Multidimensional Weibull Many others MG Lipsett, Probabilistic Simulations In this course, we have studied model-based deterministic simulations of systems In reality, a system will have uncertainties In particular, the inputs to the system may be somewhat random Noise in a signal Belgian Blocks: automobile suspensions Flow-induced random vibrations of heat exchanger piping Irregularities in raw materials into a manufacturing process Random changes in financial conditions (such as interest rate) Random delays in a supply chain MG Lipsett,
13 Randomness in Technological Systems Some continuous probability distributions in manufacturing include: Interarrival times of orders, parts, or materials Loading and unloading times Processing times (including manual tasks such as inspection of products) Time for a machine to break under a defined operating condition Time for Initial set-up of a production line or to change a machine to do a different task MG Lipsett, Different Distributions for Different Processes Failure rates of machines often follow a gamma distribution, which has a lot of flexibility and a mean that is simpler to compute than the more general Weibull distribution Interarrival times often follow an exponential distribution (based on a Poisson process) If you are not sure what distribution to use, try to fit some data to a distribution If no data are available, then you will have to assume a distribution Look at the literature for similar types of processes MG Lipsett,
14 Estimating Probabilities and Loss Functions There are different places to get data related to probability distributions: Actuarial data (other systems) Empirical data (plant data, experiment) Probabilistic transformations (Jacobian) from other PDFs through system model MATLAB Statistics Toolbox can find the best fit of a number of different types of PDFs to a data set (including combinations) For example, normfit can be used for a single-variable (univariate) Normal (Gaussian) distribution MG Lipsett, Monte Carlo Simulation Monte Carlo simulation is a probabilistic analysis of a system Inputs are probabilities (distributions or functions) Process is the system model (differential equations or discrete-events) Output is the system response (or analysis result) Solution must be robust to the bandwidth of the inputs Output Inputs Process MG Lipsett,
15 Monte Carlo Method Monte Carlo simulation can be applied to deterministic problems (such as evaluating intractable integrals) More often, Monte Carlo simulation is used for probabilistic simulations, to show how random inputs result in random outputs, thereby revealing the system behaviour. This is like finding the frequency response function for a system by driving it with a Gaussian input which has equal signal power at all frequencies, thereby generating system output at all frequencies at which the system can respond. The key is to have a good idea of the nature of the probability distributions of each of the system inputs distribution (as discussed above) MG Lipsett, Monte Carlo with a Spreadsheet As we have seen, simple Monte Carlo Discrete-event simulation can be done using a spreadsheet Excel has a random-number generator and the ability to generate values from some well-known probability distributions, as well as the ability to generate summary statistics and charts for the outputs, subject to the following limitations: Simple data structures Calculations in cells cannot be very complex Number of iterations and number of variables must be fairly small so that solution time is reasonable and storage requirements are not excessive Lab #10 has a comprehensive example of using Excel for a simple queuing problem MG Lipsett,
16 Validating a Model Validation is the process of ensuring that the simulation is solving the model correctly The simulation must be solved in such a way that the simulation responds the way the system of interest would This is of particular concern when solving a set of stiff differential equations Validation relies on having a good understanding of what an actual system would do, which usually means having some data from the system of interest (or data for known elements that will be assembled for a new system) MG Lipsett, The Trouble with Data There are some potential problems with data: The data may not be representative of what you want to model: for example, a nuclear plant in normal operating conditions may not tell you the kind of system behaviour during an emergency. Data are in the wrong format or type, compared to how data would be used in the simulation The measurement method may be prone to error, methods of recording data may have errors, and data records may have round-off errors (e.g. resolution of scales) Data may be biased due to personal interpretation or bias Data may have the wrong units (short tons, long tonnes, metric tons ) MG Lipsett,
17 More Model Validation Issues Data errors fall into two types: Precision (how tight the measurements are) Accuracy (how close the measurements are to the actual) Depending on the system, data may not be collected under the same conditions, which affects reproducibility (and variance) Invasiveness means that the act of collecting data affects the system (there s noting like having a person with a clipboard watching you to improve your performance!) Temporal and spatial averaging may add a bias problem. Observer bias is an error that prevents someone from seeing a situation or a person objectively (such biases may be subtle and hard to detect!!) MG Lipsett, Non-linear LS Estimation Linear regression deals with fitting an observed set of data to a linear equation (i.e. y = ax + b). This method assumes that each observation can be described by the model fy is a linear combination of parameters of the form fy = x i1 θ 1 + x i2 θ 2 + x i3 θ 3 + where e is a random error with mean zero and variance. In this situation the parameters a and b (θ) can be solved for explicitly using the method of least squares. Many phenomena are described by nonlinear equations where the response is a nonlinear function of the independent variable(s) and parameter(s), making it difficult or impossible to solve for parameters explicitly. MG Lipsett,
18 MATLAB Implementation of NL-LSELSE In this (more general) case it is assumed that each observation can be described by the model but this model is not a linear function of the parameters θ The method of least squares is typically used to fit a set of data to this model. However, in many cases numerical methods must be employed in order to find the set of parameters that minimizes the residual. Fortunately, MATLAB has some built in tools to perform the numerical methods needed to solve nonlinear regression problems. The one presented here is nlinfit. MG Lipsett, MATLAB nlinfit Function MG Lipsett, nlinfit is used to find least-squares parameter estimates for nonlinear models. It uses the Gauss-Newton algorithm with Levenberg-Marquardt modifications for global convergence. This function is called as follows: >> beta = nlinfit(x,y,function,beta0) where beta is a vector of the parameters that is the solution to the nonlinear regression problem X is an n x p matrix of p predictors at each of n observations y is an n x 1 vector of observed responses function is a function handle to a function of the form yhat = modelfun(beta_hat,x) beta0 is an initial prediction of the parameters 18
19 Example (from MATLAB help file): The Hougen-Watson model for reaction kinetics is an example of a nonlinear model for which one may wish to estimate the parameters from an observed experiment. The form of this model is x β 3 1x2 β5 y = 1+ β x + β x + β x Given the data from an experiment, the unknown parameters β 1 β 5 can be found using MATLAB s nlinfit function as described in the next slides MG Lipsett, Example (2) First: the predictors should be in the form of an n x p matrix of p predictors at each of n observations. For example, for p=3 predictors x1, x2, x3: x1 x2 x3 Observation 1 x11 X12 x13 Observation 2 x21 x22 x23 Observation n xn1 xn2 xn3 Or in MATLAB >> X = [x11 x12 x13; x21 x22 x23; ; xn1 xn2 xn3]; MG Lipsett,
20 Example (3) Second: the observations should be in the form of an n x 1 vector In MATLAB >> observed_rates = [rate1; rate2; ; raten]; MG Lipsett, Example (4) Third: create an m-file to evaluate the rate given the predictors and an estimate of the parameters. function rate = hougen(beta,x) b1 = beta(1); b2 = beta(2); b3 = beta(3); b4 = beta(4); b5 = beta(5); x1 = X(:,1); x2 = X(:,2); x3 = X(:,3); rate = (b1*x2 - x3/b5)./(1+b2*x1+b3*x2+b4*x3); MG Lipsett,
21 Example (5) Fourth: call nlinfit to fit the model to the data A nonlinear estimation method searches for the solution, and so it requires an initial guess beta0 for the parameters beta: >>beta0 = [beta1; beta2 beta3; beta4; beta5]; You must ensure that the MATLAB current directory is set to the location where you saved the m-file with the function describing the equation of interest (here, hougen.m ) Then you call the nlinfit function: >> beta = nlinfit(x,observed_rates,@hougen,beta0) The vector beta will contain the best fit of the nonlinear equation in terms of the parameters and the data (predictors at different observations) MG Lipsett, Estimating Parameters of System Elements If it is possible to take a system apart and measure inputs and outputs for an element of a system, then it is straightforward to estimate the functional relationship between them (which, of course, is the constitutive relationship of the element) For example, a spring can be mounted in a testing machine, and the displacement measured at different forces applied to the element Note: if the element can exhibit any nonlinearities, then we have to have data across the entire range of inputs and outputs to characterise the parameters of the constitutive relationship completely Then, we simply plug the value of each parameter into the governing equations MG Lipsett,
22 Estimating Parameters (2) In most cases, we do not have the luxury of deconstructing the system, and we have to collect data from the system and attempt to estimate all of the parameters of the system at once. MG Lipsett, Summary We use models to give us predictive capability as to how a system may behave in some circumstances This understanding allows us to make better decisions than if we had to rely on our internal models of how a system may behave Good models need to be verified (and hopefully validated) For DES and lumped-parameter systems, parameter estimation gives us increased confidence; but the estimates depend on having good data for the system elements or events MG Lipsett,
Advanced Machine Learning Practical 4b Solution: Regression (BLR, GPR & Gradient Boosting)
Advanced Machine Learning Practical 4b Solution: Regression (BLR, GPR & Gradient Boosting) Professor: Aude Billard Assistants: Nadia Figueroa, Ilaria Lauzana and Brice Platerrier E-mails: aude.billard@epfl.ch,
More information14 Random Variables and Simulation
14 Random Variables and Simulation In this lecture note we consider the relationship between random variables and simulation models. Random variables play two important roles in simulation models. We assume
More informationLecture : Probabilistic Machine Learning
Lecture : Probabilistic Machine Learning Riashat Islam Reasoning and Learning Lab McGill University September 11, 2018 ML : Many Methods with Many Links Modelling Views of Machine Learning Machine Learning
More informationTreatment of Error in Experimental Measurements
in Experimental Measurements All measurements contain error. An experiment is truly incomplete without an evaluation of the amount of error in the results. In this course, you will learn to use some common
More informationSTA 4273H: Sta-s-cal Machine Learning
STA 4273H: Sta-s-cal Machine Learning Russ Salakhutdinov Department of Computer Science! Department of Statistical Sciences! rsalakhu@cs.toronto.edu! h0p://www.cs.utoronto.ca/~rsalakhu/ Lecture 2 In our
More information6.867 Machine Learning
6.867 Machine Learning Problem set 1 Solutions Thursday, September 19 What and how to turn in? Turn in short written answers to the questions explicitly stated, and when requested to explain or prove.
More informationLecture 5. G. Cowan Lectures on Statistical Data Analysis Lecture 5 page 1
Lecture 5 1 Probability (90 min.) Definition, Bayes theorem, probability densities and their properties, catalogue of pdfs, Monte Carlo 2 Statistical tests (90 min.) general concepts, test statistics,
More informationSYMBIOSIS CENTRE FOR DISTANCE LEARNING (SCDL) Subject: production and operations management
Sample Questions: Section I: Subjective Questions 1. What are the inputs required to plan a master production schedule? 2. What are the different operations schedule types based on time and applications?
More informationISE/OR 762 Stochastic Simulation Techniques
ISE/OR 762 Stochastic Simulation Techniques Topic 0: Introduction to Discrete Event Simulation Yunan Liu Department of Industrial and Systems Engineering NC State University January 9, 2018 Yunan Liu (NC
More informationNATCOR. Forecast Evaluation. Forecasting with ARIMA models. Nikolaos Kourentzes
NATCOR Forecast Evaluation Forecasting with ARIMA models Nikolaos Kourentzes n.kourentzes@lancaster.ac.uk O u t l i n e 1. Bias measures 2. Accuracy measures 3. Evaluation schemes 4. Prediction intervals
More informationLecture - 02 Rules for Pinch Design Method (PEM) - Part 02
Process Integration Prof. Bikash Mohanty Department of Chemical Engineering Indian Institute of Technology, Roorkee Module - 05 Pinch Design Method for HEN synthesis Lecture - 02 Rules for Pinch Design
More information6.867 Machine Learning
6.867 Machine Learning Problem set 1 Due Thursday, September 19, in class What and how to turn in? Turn in short written answers to the questions explicitly stated, and when requested to explain or prove.
More informationA Scientific Model for Free Fall.
A Scientific Model for Free Fall. I. Overview. This lab explores the framework of the scientific method. The phenomenon studied is the free fall of an object released from rest at a height H from the ground.
More informationSECTION 7: CURVE FITTING. MAE 4020/5020 Numerical Methods with MATLAB
SECTION 7: CURVE FITTING MAE 4020/5020 Numerical Methods with MATLAB 2 Introduction Curve Fitting 3 Often have data,, that is a function of some independent variable,, but the underlying relationship is
More informationPredicting AGI: What can we say when we know so little?
Predicting AGI: What can we say when we know so little? Fallenstein, Benja Mennen, Alex December 2, 2013 (Working Paper) 1 Time to taxi Our situation now looks fairly similar to our situation 20 years
More informationGaussian processes. Chuong B. Do (updated by Honglak Lee) November 22, 2008
Gaussian processes Chuong B Do (updated by Honglak Lee) November 22, 2008 Many of the classical machine learning algorithms that we talked about during the first half of this course fit the following pattern:
More informationChapter 5: Forecasting
1 Textbook: pp. 165-202 Chapter 5: Forecasting Every day, managers make decisions without knowing what will happen in the future 2 Learning Objectives After completing this chapter, students will be able
More informationBayesian Machine Learning
Bayesian Machine Learning Andrew Gordon Wilson ORIE 6741 Lecture 2: Bayesian Basics https://people.orie.cornell.edu/andrew/orie6741 Cornell University August 25, 2016 1 / 17 Canonical Machine Learning
More informationIndustrial Engineering Prof. Inderdeep Singh Department of Mechanical & Industrial Engineering Indian Institute of Technology, Roorkee
Industrial Engineering Prof. Inderdeep Singh Department of Mechanical & Industrial Engineering Indian Institute of Technology, Roorkee Module - 04 Lecture - 05 Sales Forecasting - II A very warm welcome
More informationChapter 7 (Cont d) PERT
Chapter 7 (Cont d) PERT Project Management for Business, Engineering, and Technology Prepared by John Nicholas, Ph.D. Loyola University Chicago & Herman Steyn, PhD University of Pretoria Variability of
More informationSYDE 372 Introduction to Pattern Recognition. Probability Measures for Classification: Part I
SYDE 372 Introduction to Pattern Recognition Probability Measures for Classification: Part I Alexander Wong Department of Systems Design Engineering University of Waterloo Outline 1 2 3 4 Why use probability
More informationBayesian Methods for Machine Learning
Bayesian Methods for Machine Learning CS 584: Big Data Analytics Material adapted from Radford Neal s tutorial (http://ftp.cs.utoronto.ca/pub/radford/bayes-tut.pdf), Zoubin Ghahramni (http://hunch.net/~coms-4771/zoubin_ghahramani_bayesian_learning.pdf),
More informationSequential Decision Problems
Sequential Decision Problems Michael A. Goodrich November 10, 2006 If I make changes to these notes after they are posted and if these changes are important (beyond cosmetic), the changes will highlighted
More informationModel-building and parameter estimation
Luleå University of Technology Johan Carlson Last revision: July 27, 2009 Measurement Technology and Uncertainty Analysis - E7021E MATLAB homework assignment Model-building and parameter estimation Introduction
More informationDiscrete-event simulations
Discrete-event simulations Lecturer: Dmitri A. Moltchanov E-mail: moltchan@cs.tut.fi http://www.cs.tut.fi/kurssit/elt-53606/ OUTLINE: Why do we need simulations? Step-by-step simulations; Classifications;
More informationRandomized Algorithms
Randomized Algorithms Prof. Tapio Elomaa tapio.elomaa@tut.fi Course Basics A new 4 credit unit course Part of Theoretical Computer Science courses at the Department of Mathematics There will be 4 hours
More informationSTA441: Spring Multiple Regression. More than one explanatory variable at the same time
STA441: Spring 2016 Multiple Regression More than one explanatory variable at the same time This slide show is a free open source document. See the last slide for copyright information. One Explanatory
More informationCS 542G: Robustifying Newton, Constraints, Nonlinear Least Squares
CS 542G: Robustifying Newton, Constraints, Nonlinear Least Squares Robert Bridson October 29, 2008 1 Hessian Problems in Newton Last time we fixed one of plain Newton s problems by introducing line search
More informationGeneralized Linear Models for Non-Normal Data
Generalized Linear Models for Non-Normal Data Today s Class: 3 parts of a generalized model Models for binary outcomes Complications for generalized multivariate or multilevel models SPLH 861: Lecture
More informationChapter 13: Forecasting
Chapter 13: Forecasting Assistant Prof. Abed Schokry Operations and Productions Management First Semester 2013-2014 Chapter 13: Learning Outcomes You should be able to: List the elements of a good forecast
More informationA First Course on Kinetics and Reaction Engineering Supplemental Unit S4. Numerically Fitting Models to Data
Supplemental Unit S4. Numerically Fitting Models to Data Defining the Problem Many software routines for fitting a model to experimental data can only be used when the model is of a pre-defined mathematical
More informationPractical Statistics
Practical Statistics Lecture 1 (Nov. 9): - Correlation - Hypothesis Testing Lecture 2 (Nov. 16): - Error Estimation - Bayesian Analysis - Rejecting Outliers Lecture 3 (Nov. 18) - Monte Carlo Modeling -
More informationDepartment of Electrical- and Information Technology. ETS061 Lecture 3, Verification, Validation and Input
ETS061 Lecture 3, Verification, Validation and Input Verification and Validation Real system Validation Model Verification Measurements Verification Break model down into smaller bits and test each bit
More informationStatistical Methods in Particle Physics
Statistical Methods in Particle Physics Lecture 11 January 7, 2013 Silvia Masciocchi, GSI Darmstadt s.masciocchi@gsi.de Winter Semester 2012 / 13 Outline How to communicate the statistical uncertainty
More informationAlgorithm Independent Topics Lecture 6
Algorithm Independent Topics Lecture 6 Jason Corso SUNY at Buffalo Feb. 23 2009 J. Corso (SUNY at Buffalo) Algorithm Independent Topics Lecture 6 Feb. 23 2009 1 / 45 Introduction Now that we ve built an
More informationUncertainty and Graphical Analysis
Uncertainty and Graphical Analysis Introduction Two measures of the quality of an experimental result are its accuracy and its precision. An accurate result is consistent with some ideal, true value, perhaps
More informationAUTO SALES FORECASTING FOR PRODUCTION PLANNING AT FORD
FCAS AUTO SALES FORECASTING FOR PRODUCTION PLANNING AT FORD Group - A10 Group Members: PGID Name of the Member 1. 61710956 Abhishek Gore 2. 61710521 Ajay Ballapale 3. 61710106 Bhushan Goyal 4. 61710397
More informationMachine Learning CSE546 Sham Kakade University of Washington. Oct 4, What about continuous variables?
Linear Regression Machine Learning CSE546 Sham Kakade University of Washington Oct 4, 2016 1 What about continuous variables? Billionaire says: If I am measuring a continuous variable, what can you do
More informationWarwick Business School Forecasting System. Summary. Ana Galvao, Anthony Garratt and James Mitchell November, 2014
Warwick Business School Forecasting System Summary Ana Galvao, Anthony Garratt and James Mitchell November, 21 The main objective of the Warwick Business School Forecasting System is to provide competitive
More informationEAS 535 Laboratory Exercise Weather Station Setup and Verification
EAS 535 Laboratory Exercise Weather Station Setup and Verification Lab Objectives: In this lab exercise, you are going to examine and describe the error characteristics of several instruments, all purportedly
More informationAssistant Prof. Abed Schokry. Operations and Productions Management. First Semester
Chapter 3 Forecasting Assistant Prof. Abed Schokry Operations and Productions Management First Semester 2010 2011 Chapter 3: Learning Outcomes You should be able to: List the elements of a good forecast
More informationHuman-Oriented Robotics. Probability Refresher. Kai Arras Social Robotics Lab, University of Freiburg Winter term 2014/2015
Probability Refresher Kai Arras, University of Freiburg Winter term 2014/2015 Probability Refresher Introduction to Probability Random variables Joint distribution Marginalization Conditional probability
More informationMachine Learning CSE546 Carlos Guestrin University of Washington. September 30, What about continuous variables?
Linear Regression Machine Learning CSE546 Carlos Guestrin University of Washington September 30, 2014 1 What about continuous variables? n Billionaire says: If I am measuring a continuous variable, what
More informationSeminar on Case Studies in Operations Research (Mat )
Seminar on Case Studies in Operations Research (Mat-2.4177) Evidential Uncertainties in Reliability Assessment - Study of Non-Destructive Testing of Final Disposal Canisters VTT, Posiva Backlund, Ville-Pekka
More informationExercises Tutorial at ICASSP 2016 Learning Nonlinear Dynamical Models Using Particle Filters
Exercises Tutorial at ICASSP 216 Learning Nonlinear Dynamical Models Using Particle Filters Andreas Svensson, Johan Dahlin and Thomas B. Schön March 18, 216 Good luck! 1 [Bootstrap particle filter for
More informationLogistic Regression: Regression with a Binary Dependent Variable
Logistic Regression: Regression with a Binary Dependent Variable LEARNING OBJECTIVES Upon completing this chapter, you should be able to do the following: State the circumstances under which logistic regression
More informationStructural Reliability
Structural Reliability Thuong Van DANG May 28, 2018 1 / 41 2 / 41 Introduction to Structural Reliability Concept of Limit State and Reliability Review of Probability Theory First Order Second Moment Method
More information11/8/2018. Overview. PERT / CPM Part 2
/8/08 PERT / CPM Part BSAD 0 Dave Novak Fall 08 Source: Anderson et al., 0 Quantitative Methods for Business th edition some slides are directly from J. Loucks 0 Cengage Learning Overview Last class introduce
More informationCS6375: Machine Learning Gautam Kunapuli. Decision Trees
Gautam Kunapuli Example: Restaurant Recommendation Example: Develop a model to recommend restaurants to users depending on their past dining experiences. Here, the features are cost (x ) and the user s
More informationConditional Distribution Fitting of High Dimensional Stationary Data
Conditional Distribution Fitting of High Dimensional Stationary Data Miguel Cuba and Oy Leuangthong The second order stationary assumption implies the spatial variability defined by the variogram is constant
More information1 Measurement Uncertainties
1 Measurement Uncertainties (Adapted stolen, really from work by Amin Jaziri) 1.1 Introduction No measurement can be perfectly certain. No measuring device is infinitely sensitive or infinitely precise.
More informationChapter 8 Statistical Quality Control, 7th Edition by Douglas C. Montgomery. Copyright (c) 2013 John Wiley & Sons, Inc.
1 Learning Objectives Chapter 8 Statistical Quality Control, 7th Edition by Douglas C. Montgomery. 2 Process Capability Natural tolerance limits are defined as follows: Chapter 8 Statistical Quality Control,
More informationL11: Pattern recognition principles
L11: Pattern recognition principles Bayesian decision theory Statistical classifiers Dimensionality reduction Clustering This lecture is partly based on [Huang, Acero and Hon, 2001, ch. 4] Introduction
More informationCS 147: Computer Systems Performance Analysis
CS 147: Computer Systems Performance Analysis Advanced Regression Techniques CS 147: Computer Systems Performance Analysis Advanced Regression Techniques 1 / 31 Overview Overview Overview Common Transformations
More informationCourse Project. Physics I with Lab
COURSE OBJECTIVES 1. Explain the fundamental laws of physics in both written and equation form 2. Describe the principles of motion, force, and energy 3. Predict the motion and behavior of objects based
More informationRatio of Polynomials Fit One Variable
Chapter 375 Ratio of Polynomials Fit One Variable Introduction This program fits a model that is the ratio of two polynomials of up to fifth order. Examples of this type of model are: and Y = A0 + A1 X
More informationQuiz 1. Name: Instructions: Closed book, notes, and no electronic devices.
Quiz 1. Name: Instructions: Closed book, notes, and no electronic devices. 1. What is the difference between a deterministic model and a probabilistic model? (Two or three sentences only). 2. What is the
More informationQueens College, CUNY, Department of Computer Science Numerical Methods CSCI 361 / 761 Spring 2018 Instructor: Dr. Sateesh Mane.
Queens College, CUNY, Department of Computer Science Numerical Methods CSCI 361 / 761 Spring 2018 Instructor: Dr. Sateesh Mane c Sateesh R. Mane 2018 3 Lecture 3 3.1 General remarks March 4, 2018 This
More informationRequirements Validation. Content. What the standards say (*) ?? Validation, Verification, Accreditation!! Correctness and completeness
Requirements Validation Requirements Management Requirements Validation?? Validation, Verification, Accreditation!! Check if evrything is OK With respect to what? Mesurement associated with requirements
More information1 Basic Analysis of Forward-Looking Decision Making
1 Basic Analysis of Forward-Looking Decision Making Individuals and families make the key decisions that determine the future of the economy. The decisions involve balancing current sacrifice against future
More informationStep 1 Determine the Order of the Reaction
Step 1 Determine the Order of the Reaction In this step, you fit the collected data to the various integrated rate expressions for zezro-, half-, first-, and second-order kinetics. (Since the reaction
More informationOverfitting, Bias / Variance Analysis
Overfitting, Bias / Variance Analysis Professor Ameet Talwalkar Professor Ameet Talwalkar CS260 Machine Learning Algorithms February 8, 207 / 40 Outline Administration 2 Review of last lecture 3 Basic
More informationStatistics and Data Analysis
Statistics and Data Analysis The Crash Course Physics 226, Fall 2013 "There are three kinds of lies: lies, damned lies, and statistics. Mark Twain, allegedly after Benjamin Disraeli Statistics and Data
More informationExcel for Scientists and Engineers Numerical Method s. E. Joseph Billo
Excel for Scientists and Engineers Numerical Method s E. Joseph Billo Detailed Table of Contents Preface Acknowledgments About the Author Chapter 1 Introducing Visual Basic for Applications 1 Chapter
More information1 Using standard errors when comparing estimated values
MLPR Assignment Part : General comments Below are comments on some recurring issues I came across when marking the second part of the assignment, which I thought it would help to explain in more detail
More informationApplied Computational Economics Workshop. Part 3: Nonlinear Equations
Applied Computational Economics Workshop Part 3: Nonlinear Equations 1 Overview Introduction Function iteration Newton s method Quasi-Newton methods Practical example Practical issues 2 Introduction Nonlinear
More informationSTAT 509 Section 3.4: Continuous Distributions. Probability distributions are used a bit differently for continuous r.v. s than for discrete r.v. s.
STAT 509 Section 3.4: Continuous Distributions Probability distributions are used a bit differently for continuous r.v. s than for discrete r.v. s. A continuous random variable is one for which the outcome
More informationLecture 2. Judging the Performance of Classifiers. Nitin R. Patel
Lecture 2 Judging the Performance of Classifiers Nitin R. Patel 1 In this note we will examine the question of how to udge the usefulness of a classifier and how to compare different classifiers. Not only
More informationReview of the role of uncertainties in room acoustics
Review of the role of uncertainties in room acoustics Ralph T. Muehleisen, Ph.D. PE, FASA, INCE Board Certified Principal Building Scientist and BEDTR Technical Lead Division of Decision and Information
More informationParameter Estimation. Industrial AI Lab.
Parameter Estimation Industrial AI Lab. Generative Model X Y w y = ω T x + ε ε~n(0, σ 2 ) σ 2 2 Maximum Likelihood Estimation (MLE) Estimate parameters θ ω, σ 2 given a generative model Given observed
More information. Introduction to CPM / PERT Techniques. Applications of CPM / PERT. Basic Steps in PERT / CPM. Frame work of PERT/CPM. Network Diagram Representation. Rules for Drawing Network Diagrams. Common Errors
More informationMachine Learning CSE546 Carlos Guestrin University of Washington. September 30, 2013
Bayesian Methods Machine Learning CSE546 Carlos Guestrin University of Washington September 30, 2013 1 What about prior n Billionaire says: Wait, I know that the thumbtack is close to 50-50. What can you
More informationIntroduction to ecosystem modelling Stages of the modelling process
NGEN02 Ecosystem Modelling 2018 Introduction to ecosystem modelling Stages of the modelling process Recommended reading: Smith & Smith Environmental Modelling, Chapter 2 Models in science and research
More informationAdvances in promotional modelling and analytics
Advances in promotional modelling and analytics High School of Economics St. Petersburg 25 May 2016 Nikolaos Kourentzes n.kourentzes@lancaster.ac.uk O u t l i n e 1. What is forecasting? 2. Forecasting,
More informationStatistics: Learning models from data
DS-GA 1002 Lecture notes 5 October 19, 2015 Statistics: Learning models from data Learning models from data that are assumed to be generated probabilistically from a certain unknown distribution is a crucial
More informationCS 700: Quantitative Methods & Experimental Design in Computer Science
CS 700: Quantitative Methods & Experimental Design in Computer Science Sanjeev Setia Dept of Computer Science George Mason University Logistics Grade: 35% project, 25% Homework assignments 20% midterm,
More informationBig Data Analysis with Apache Spark UC#BERKELEY
Big Data Analysis with Apache Spark UC#BERKELEY This Lecture: Relation between Variables An association A trend» Positive association or Negative association A pattern» Could be any discernible shape»
More informationA particularly nasty aspect of this is that it is often difficult or impossible to tell if a model fails to satisfy these steps.
ECON 497: Lecture 6 Page 1 of 1 Metropolitan State University ECON 497: Research and Forecasting Lecture Notes 6 Specification: Choosing the Independent Variables Studenmund Chapter 6 Before we start,
More informationPhysics 509: Bootstrap and Robust Parameter Estimation
Physics 509: Bootstrap and Robust Parameter Estimation Scott Oser Lecture #20 Physics 509 1 Nonparametric parameter estimation Question: what error estimate should you assign to the slope and intercept
More informationThe Impact of Distributed Generation on Power Transmission Grid Dynamics
The Impact of Distributed Generation on Power Transmission Grid Dynamics D. E. Newman B. A. Carreras M. Kirchner I. Dobson Physics Dept. University of Alaska Fairbanks AK 99775 Depart. Fisica Universidad
More informationOverview. Probabilistic Interpretation of Linear Regression Maximum Likelihood Estimation Bayesian Estimation MAP Estimation
Overview Probabilistic Interpretation of Linear Regression Maximum Likelihood Estimation Bayesian Estimation MAP Estimation Probabilistic Interpretation: Linear Regression Assume output y is generated
More informationCSC2515 Assignment #2
CSC2515 Assignment #2 Due: Nov.4, 2pm at the START of class Worth: 18% Late assignments not accepted. 1 Pseudo-Bayesian Linear Regression (3%) In this question you will dabble in Bayesian statistics and
More informationGaussians Linear Regression Bias-Variance Tradeoff
Readings listed in class website Gaussians Linear Regression Bias-Variance Tradeoff Machine Learning 10701/15781 Carlos Guestrin Carnegie Mellon University January 22 nd, 2007 Maximum Likelihood Estimation
More informationIntroduction to ecosystem modelling (continued)
NGEN02 Ecosystem Modelling 2015 Introduction to ecosystem modelling (continued) Uses of models in science and research System dynamics modelling The modelling process Recommended reading: Smith & Smith
More informationExploratory Factor Analysis and Principal Component Analysis
Exploratory Factor Analysis and Principal Component Analysis Today s Topics: What are EFA and PCA for? Planning a factor analytic study Analysis steps: Extraction methods How many factors Rotation and
More informationPart 4: Multi-parameter and normal models
Part 4: Multi-parameter and normal models 1 The normal model Perhaps the most useful (or utilized) probability model for data analysis is the normal distribution There are several reasons for this, e.g.,
More informationModeling and Performance Analysis with Discrete-Event Simulation
Simulation Modeling and Performance Analysis with Discrete-Event Simulation Chapter 9 Input Modeling Contents Data Collection Identifying the Distribution with Data Parameter Estimation Goodness-of-Fit
More informationStructural Uncertainty in Health Economic Decision Models
Structural Uncertainty in Health Economic Decision Models Mark Strong 1, Hazel Pilgrim 1, Jeremy Oakley 2, Jim Chilcott 1 December 2009 1. School of Health and Related Research, University of Sheffield,
More informationComputer Science, Informatik 4 Communication and Distributed Systems. Simulation. Discrete-Event System Simulation. Dr.
Simulation Discrete-Event System Simulation Chapter 8 Input Modeling Purpose & Overview Input models provide the driving force for a simulation model. The quality of the output is no better than the quality
More informationUsing Microsoft Excel
Using Microsoft Excel Objective: Students will gain familiarity with using Excel to record data, display data properly, use built-in formulae to do calculations, and plot and fit data with linear functions.
More informationDecision 411: Class 3
Decision 411: Class 3 Discussion of HW#1 Introduction to seasonal models Seasonal decomposition Seasonal adjustment on a spreadsheet Forecasting with seasonal adjustment Forecasting inflation Poor man
More informationBusiness Statistics. Lecture 9: Simple Regression
Business Statistics Lecture 9: Simple Regression 1 On to Model Building! Up to now, class was about descriptive and inferential statistics Numerical and graphical summaries of data Confidence intervals
More informationQuadratic Equations Part I
Quadratic Equations Part I Before proceeding with this section we should note that the topic of solving quadratic equations will be covered in two sections. This is done for the benefit of those viewing
More informationExpected Value II. 1 The Expected Number of Events that Happen
6.042/18.062J Mathematics for Computer Science December 5, 2006 Tom Leighton and Ronitt Rubinfeld Lecture Notes Expected Value II 1 The Expected Number of Events that Happen Last week we concluded by showing
More informationBasics of Uncertainty Analysis
Basics of Uncertainty Analysis Chapter Six Basics of Uncertainty Analysis 6.1 Introduction As shown in Fig. 6.1, analysis models are used to predict the performances or behaviors of a product under design.
More informationStatistics for scientists and engineers
Statistics for scientists and engineers February 0, 006 Contents Introduction. Motivation - why study statistics?................................... Examples..................................................3
More informationEvaluating the value of structural heath monitoring with longitudinal performance indicators and hazard functions using Bayesian dynamic predictions
Evaluating the value of structural heath monitoring with longitudinal performance indicators and hazard functions using Bayesian dynamic predictions C. Xing, R. Caspeele, L. Taerwe Ghent University, Department
More informationExploratory Factor Analysis and Principal Component Analysis
Exploratory Factor Analysis and Principal Component Analysis Today s Topics: What are EFA and PCA for? Planning a factor analytic study Analysis steps: Extraction methods How many factors Rotation and
More informationTrendlines Simple Linear Regression Multiple Linear Regression Systematic Model Building Practical Issues
Trendlines Simple Linear Regression Multiple Linear Regression Systematic Model Building Practical Issues Overfitting Categorical Variables Interaction Terms Non-linear Terms Linear Logarithmic y = a +
More information+ + ( + ) = Linear recurrent networks. Simpler, much more amenable to analytic treatment E.g. by choosing
Linear recurrent networks Simpler, much more amenable to analytic treatment E.g. by choosing + ( + ) = Firing rates can be negative Approximates dynamics around fixed point Approximation often reasonable
More information