Study Guides (248,280)
BU275 (19)
Final

# 2012 BDM BU275 Final Exam Cheatsheet

2 Pages
702 Views

Department
Course
BU275
Professor
Michael Pavlin
Semester
Winter

Description
Probability Distributions: is < or > then 2 column DECISIONS WITH UNCERTAINTY Optimal decision under risk = D1 Uniform: each result has same probability Anderson-Darling Laplace (assume equally likely) EOL(D1) = .6(0) + .2(0) +.2(9000) = 1800 (rolling a single dice) Modification of K-S, placing greater weight MAXIMAX/MINIMAX (optimistic) Opportunity loss matrix Normal: z is the number of stan dev’s away on the tails of the distribution, value < 1.5 MAXIMIN / MINIMAX (conservative) from the mean = z = value – mean / st dev. indicates good fit MINIMAX REGRET S1 S2 S3 Binomial: (put formula here) If a decision is dominated in all D1 0 0 9000 Triangular: has mode (most likely), max and outcomes, then you can remove it. D2 2000 500 4000 min. LAPLACE (choose higher Poisson: a discrete distribution (a count) average/total payoff) D3 5000 1500 0 number of successes in an interval (λ) x − λ depends on the success λ e rate and the interval length.x) = f(x) = x! Exponential: μ is average time to complete a task, or interval between events. It is a continuous distribution. If poisson rate applies, then interval between is exponential. P x≤x =1−e ) (−x0μ ) Compare the KS / AD values to the chart or MAXIMAX (MAX the MAXimum Payoff) Above formula will give0you percentage that ‘good fit’ indicator. Take the Maximum value from all the maxs UTILITY the task will be completed within x mins of Utilities are used when the decision the average.Turn the interval (μ) into a rate Validation and Verification criteria must be based on more than just  λ = 1/μ – f(x) = λe λx Verification done by expert, validation expected monetary values. can be done by using past data in Utility measures of the value of a Simulation is the process of designing a particular outcome, reflecting the simulation and seeing if results are correct. mathematical or logical model of a real Steps in Validation: decision maker’s attitude towards a system and then conducting computer-based 1. develop model that looks reasonable to collection of factors. experiments with the model to describe, those that understand system MAXIMAX (MAX the Minimum Payoff) These factors may be profit, loss, and risk. explain, and predict the behavior of the real 2. Validate assumption Take the greatest value from all the mins Deriving Utility system. Simulation is the most widely 3. Validate output Assign 100 utiles and 0 utiles to best and used Decision Analysis technique used to You are looking for the long run steady worst case payoffs. gain knowledge about outcomes to assess Take value in middle, and give them that state operation of a system. risk for certain or the gamble to get best or Continuous Simulation: based on math eqn’s QUEUING worst. The probability of getting the best and used for simulating continuous values. Size of population can be infinite or finite that makes these 2 options equal is the utility of the middle value. (time spend waiting in line) Behaviour of arrivals (patient, balk, renege) Discrete Simulation: simulation specific Waiting line can have limited or unlimited values or points (number waiting in queue) capacity MINIMAX Regret (minimize Opport Cost) Advan: leads to better understanding, can Take the oppourtunity loss of each cell, and RISK AVERSE X/X/# model any assumptions, easier to explain, M – Markov Distribution (Poss / expon) take the decision which has the least highest provides a more realistic replication. D – Deterministic: constant regret. Disadvan: no guarantee it will provide good G – General Distribution with known Regret States of Nature Maximum Tnable s s s Regret results, no way to prove reliable, time mean/variance i 1 2 3 consuming, it’s random based may be less λ = average arrival rate c d1 0 0 9,000 9,000 RISK LOVERS accurate. 1/λ = average time between arrivals e Static vs Dynamic μ = average service rate for each server D d 2 2,000 500 4,000 4,000 Does time have a role in model? Dynamic 1/ μ = average service time d3 5,000 1,500 0 5,000 simulation is time driven. Arrival Rate Distribution Expected Utility Approach Continuous change vs Discrete change Question: What is the probability that no For each alternative, multiply the utility orders are received within a 15- These are for PROFIT, the opposite are for by the probability for that state. Sum for Can system change continuously or only at minute period? (mean is 5 in a 15 discrete points in time COST questions. each alternative, and take highest Deterministic vs Probabilistic x - interval) alternative. Is everything certain, or is there uncertainty?x) = (λ e )/x! DECISIONS under RISK Neutral Approach uses Normal EV P(0) = (5 e )/0! Use Monte Carlo technique for uncertainty. 3 options: 1) avoid, 2) manage (insurance), 3) Doing probabilistic and taking averages over Service Time Distribution Embrace (casinos) time is better then doing best case/worst casQuestion: What % of orders will take Probabilities known for outcomes. (determ) Use Expected Value Approach and take less than one minute to Most operational models process? (mean = 30/hr, 2 min highest EV. Dynamic, Discrete change, Probabilistic an order): P(T ≤ t) = 1 - e = 1 - e0(1/60) Mapping this Utility Chart: The Monte Carlo technique is defined as a = 1 - .6065 technique for selecting numbers randomly
More Less

Related notes for BU275
Me

OR

Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Join to view

OR

By registering, I agree to the Terms and Privacy Policies
Just a few more details

So we can recommend you notes for your school.