Pre-reading Queueing

Theory
-Prof. Utsav Pandey
Probability Density/Mass Function
• For a continuous random variable, the PDF (probability density function) at
a particular value represents the relative likelihood that the random variable
will take on a value near that point.
• It shows how "dense" the probability is around that specific value.
• A higher PDF value indicates a higher probability of the variable falling close to that
point. (it never gives the probability of exact value which is always 0 for continuous
variables).
• The value of the PDF at a single point is not a probability, but
rather a measure of the relative probability density at that
point.
• The area under the PDF curve is always one.
Poisson Process
1. No. of people coming to an ATM counter.
2. No. of people born in a city in a specific time window (between T and T+).
3. No. of goals hit by a hockey player.
• Counting process which possesses the following properties
A. Independent increment: The number of events occurring in disjoint intervals are
independent. If N(T+)-N(T)=v(T), then v(10) and V(11) are independent .
B. Stationary increment: no. of incidents depends only on the length , not on the stating
time T.
• Scenario 1 follows property (A), not (B) as rush hours have larger arrival for the same
interval length.
• Scenario 2 doesn’t follow (A): Since people born in an interval is a major determinant
of the number of people born in the next interval.
• More or less follows (B), no pattern is seen where more babies are born in a particular time.
• Scenario 3 follows (A) unless we consider psychological factors. Does not follow (B)
since the player’s form might vary depending on the start time T.
Poisson Process Properties
• Assumptions:
• k is the number of times an event occurs in an interval and k can take values 0, 1, 2, ... .
• The occurrence of one event does not affect the probability that a second event will
occur. That is, events occur independently.
• The average rate at which events occur is independent of any occurrences.
• For simplicity, this is usually assumed to be constant, but may vary with time in
practice .
• Two events cannot occur at exactly the same instant; instead, at each small sub-
interval, either exactly one event occurs, or no event occurs.
 Violation of Poisson assumptions
• The number of students who arrive at the student union per minute will not follow a
Poisson distribution, because the rate is not constant
• low rate during class time, high rate between class times (non-constant arrival rate
may be modeled as a mixed Poisson distribution)
• and the arrivals of individual students are not independent
• students tend to come in groups. (The arrival of groups is a
compound Poisson process)
• Click on a FB/Insta post:
• Recency Bias: More seen/view in the beginning.
• Also during a 24-hour haul, during night or office hour the views would be less.
• The number of petrol pumps located along the Lucknow-Agra highway is not random.
As one petrol pump reduces the probability of another petrol pump in its vicinity.
• Examples in which at least one event is guaranteed are not Poisson distributed; but
may be modeled using a zero-truncated Poisson distribution.
• Count distributions in which the number of intervals with zero events is higher than
predicted by a Poisson model may be modeled using a zero-inflated model.
Possible values of X and in Poisson Distribution
• The values realized/assumed by a R.V. X following a Poisson distribution are
always discrete (not continuous).
• Note, discrete values are not same as integers. Fraction values can be discrete in nature.
• But mostly the values of “count” or “rank” are natural X=5 mean = 3
numbers-so that X has to be integer. 0.10081

• Mean can take fraction values (since it refers to the X=5 mean = 3.5
0.13216
average occurrence per unit time, if the time unit is
X=5.5 mean 3
changed an integer mean transforms into a fraction.) 0.10081

• Excel function takes floor value of a fraction, so x=5.5 mean 3.5

0.13216
POISSON.DIST(5,3,FALSE) =POISSON.DIST(5.6,3,FALSE)
• When the mean increases, a value closer to mean is more likely to occur.
POISSON.DIST(5,3.5,FALSE) >POISSON.DIST(5,3,FALSE)
 Why exponential distribution is required?
• How likely is it that one will witness some event within the next n hours or days?

• What is the probability that some event will occur in the span between 08:00 and 14:00?

• What is the probability that the event will take at least n hours to occur?
• a postal clerk spends with his or her customer. The time is known to have an exponential
distribution with the average amount of time equal to four minutes.
• The amount of time spouses shop for anniversary cards can be modeled by an exponential
distribution with the average amount of time equal to eight minutes.
• Using the information in example 1, find the probability that a clerk spends four to five
minutes with a randomly selected customer.
• Find the 50th percentile.
• The number of days ahead travelers purchase their airline tickets can be modeled by an
exponential distribution with the average amount of time equal to 15 days. Find the
probability that a traveler will purchase a ticket fewer than ten days in advance. How many
days do half of all travelers wait?
Poisson Vs. Exponential (Inverse relation)
• Poisson: Number of event occurring per unit time interval.
• Exponential: Number of time intervals between two consecutive events.
Cars/Hour passing by a KFC outlet Hours/Cars at a KFC outlet

• A unique visitor to <Edureka> website is on average: =3/hour (Poisson mean)

• Number of hours until a unique visitor visits the site:
• Is Exponential dist. Continuous?
• But, its close associate poisson is discrete?
• Where does a Poisson distribution graph peak?
• Poisson distribution has only one parameter , which is the mean.
• Naturally the most frequent value would be , i.e., the mode.
• For exponential distribution, the graph peaks at with the value of .
• The PDF is . The first order derivative: .
• The greater the value of (or ), the faster is the decay.
Exponential Distribution
• Three scenarios are addressed by exponential distribution:
• What are the chances that a particular process will be over within time x
• Cumulative probability of x:
• During peak hours at Annapurna mess, the historical waiting time in general is 5 min.
What is the probability that you will be served the meal within 2 mins?

• What are the chances that a particular process takes at least time x?

• A shoe manufacturer provides a guarantee of 1000 days on all the products. If it fails
within the guarantee period, the purchaser will receive their money back in full.
Historically their products last for 1200 days on average. How risky is it to buy shoes
from the shoe-maker?
 Cntd…
• Sometimes we are more interested in the chances of an event occurring
within a range of time.
• For example, based on the arrival of freight train / cargo planes, you are
required to plan the transportation of goods from Pantnagar.
• The historical data shows that freight train arrives once in every four hour
(and follows Poisson distribution).
• To account for overhead work, you decide to take a window of 30 mins. So,
goods are ready for shipment 30 mins prior and 30 mins after the arrival.
• How many % of time you will not miss the shipment?
•.
• Why this apparently large time window hits such a low probability?
Poisson process-deep dive
• X is the waiting time for a random event to happen. X.
• Memoryless property: No matter how long you have been waiting for
the event to happen, at every time henceforth you start afresh!
• Mathematical Expression: .
• LHS: What is the probability that you have to wait for t more minutes
given that you have already waited for s minutes?
• RHS: What is the probability that you have to wait for at least t minutes
for the event to happen?
• Where is the mathematical proof?
Application
• A study suggests that among 10 lakh financial accounts, on average 24
accounts are reported to have engaged in dubious transactions over a period
two years. You are to find out that in a small region with 1 lakh bank
accounts, what is the probability of at most four such dubious accounts?
• If the constant rate of occurrence is 24 per 10 lakh, the rate of occurrence
per 1 lakh would be _______________.
• If the constant rate of occurrence is 24 for 2 years, then the rate of
occurrence for a year would be _______________.
• For four or fewer occurrence, the expression is