Agenda
Management of Uncertainty
Management of Uncertainty
Certainty Theory
Uncertainty is represented as a degree of belief or certainty about a fact or rule, which is known as Certainty Factor (CF). Each CF can be consider as a quantifier of the belief in a hypothesis, based on an evidence. Each assertion has to be associated with a CF, degree of
Management of Uncertainty
Certainty Theory
Certainty in the premises (facts, data) Certainty in the rules Final Value: CF = threshold
Management of Uncertainty
The certainty in a disjunction depends on the strong segment. In a disjunction the CF is the maximum of the CFs. CF (A and B) = minimum [CF(A), CF(B)] CF (A or B) = maximum [CF(A), CF(B)]
Management of Uncertainty
Example: CF (stock AB is of high technology) = 90 CF (stock AB is of high demand) = 60 CF (stock AB is of high technology and stock AB is of high demand) = 60 CF (stock AB is of high technology or stock AB is of high demand) = 90 Rule1, CF = 60 X is volatile If X is of high technology and X is of high demand. CF (X is volatile) = min [CF (stock AB is of high technology) and CF (stock AB is of high demand)] * CF/100 = min [60, 90] * 60/100 = 60 * 60 / 100 = 36
Management of Uncertainty
Probabilistic Methods:
Classic: It refers to the expectations about the occurrence of an event, based on the frequency of the results of the event occurred in the past. The relative frequency of the occurrence of an event approximates to its probability.
Management of Uncertainty
Probabilities
Probability is method that allow us translate an opinion or expectation of somebody or something in numbers.
Management of Uncertainty
The probability is a number between 0 and 1; it reflects the probable occurrence of an event. An impossible event has probability 0. A certain event has probability 1. The probability of an event A is denoted by P(A) The probability of 2 events A and B, occurring at the same time is
Management of Uncertainty
The set of independent events (called the reference space, such as: A1, A2, . . . ,An), can represent all possibilities of occurrence.
The sum of all probabilities of all events must be equal to the certainty: 1.
Management of Uncertainty
Bayesian Method: Probability is a degree of belief that a person has about some hypothesis, event or uncertain quantity.
Management of Uncertainty
Bayesian Method:
This concept is based on the principle that we have to know the probability of a previous event. The theorem provides a mathematical model to reason, combining the a previous belief with the evidence to estimate the uncertainty.
It allows us to express the probability P(A | B) (post probability) in terms of the conditional probability P(B | A), P(A) y P(B).
Management of Uncertainty
In general:
P E | H i * P H i P H i | E P E | H 1 * P H 1 ... P E | H n * P H n
Management of Uncertainty
Conditional and Combined Probability
In the real world the probability of occurrence of an event is hardly difficult to occur independently of other events. However, probabilities and uncertainty must work with events that are not independent.
Management of Uncertainty
If two independent events A, B have probabilities P(A), P(B), the combined probability of both events is: P(A and B) = P(A) * P(B). P(A or B) = P(A) + P(B) - P(A) * P(B)
The conditional probability relates the probability of an event given the occurrence of another event. The estimated probability of an event will change if the context changes.
Management of Uncertainty
Example: Probabilistic Sum CF = 60 X is volatile If X is high technology or X is in high demand. CF (X is volatile) = CF * {[(CF (stock AB is of high technology ) + (CF (stock AB is of high demand)] [(CF (stock AB is of high technology ) * (CF (stock AB is of high demand)]/100} / 100 = = 60 * {[(90 + 60] [(90 * 60)/100]}/100 57.6
Rule2:
Management of Uncertainty
Example :
Determine the probability that a person is a man, knowing that his age is 80 years.
Previously we know that: The probability that a person, randomly chosen, is a man is: 0.5
The probability that a person, randomly chosen, is 80 years old is: 0.005
The probability that any person, randomly chosen, is a man and 80 years old is: 0.002
Management of Uncertainty
The conditional probability of an event A, given the occurrence of an event B is: P(A | B) = P(A and B) / P(B)
Then: P(X is man | age of X is 80) = P(X is man and age of X is 80) / P(age of X is 80) = 0.002 / 0.005 = 0.4
When a probability of an event is assigned, the conditional probability takes into consideration the particular characteristics of the settings.
Management of Uncertainty
Difficulties
Management of Uncertainty
Difficulties
1. The Bayes law requires the availability of all previous and conditional probabilities. In practice, this can be difficult to
achieve.
2. The Bayes law is mathematical correct only if all possible results do not depend upon themselves. 3. As the knowledge base grows, it gets difficult to change (update) a probability, without causing other probabilities to change, in the domain space, in order to mitigate the effects and keep the assumption that: P(H1) + P(H2) + . . . + P(Hn) = 1.