Module I: Probability Distributions

Topic 3: Poisson Distribution

Dr. P. Rajendra, Professor, Department of Mathematics,

CMRIT, Bengaluru.
Introduction to Poisson Distribution

▶ The Poisson Distribution is a discrete probability

distribution that expresses the probability of a given number
of events occurring in a fixed interval of time or space.
▶ The key assumptions are:
▶ Events occur independently of each other.
▶ The average rate (λ) at which events occur is constant.
▶ Two events cannot occur at exactly the same time.
▶ The Poisson Distribution is particularly useful for modeling
the number of times an event occurs within a specific time
period or spatial area.

Example
The Poisson Distribution can model the number of user clicks on a
webpage per minute or the number of requests to a server per hour.
Poisson Distribution (Continued..)
▶ The probability mass function (PMF) of the Poisson
Distribution is given by:

λk e −λ
P(X = k) =
k!
▶ Here:
▶ X is the random variable representing the number of events.
▶ k is the actual number of events that occur.
▶ λ is the average rate (mean number of events in the interval).
▶ e is the base of the natural logarithm (approximately 2.71828).
▶ The PMF gives the probability of observing exactly k events
in a fixed interval.

Example
Suppose a server receives an average of 5 requests per minute.
The Poisson Distribution can be used to find the probability that
the server receives exactly 8 requests in a given minute.
Poisson Distribution (Continued..)

Let X be the discrete random variable for any real value λ, such
that the probability mass function of the Poisson distribution can
be defined as:
( −λ x
e λ
P(X = x) = P(x) = x! , x ≥ 0
0, otherwise
where λ is called the parameter.
i) P(X = x) = P(x) ≥ 0
P∞ P∞ e −λ λx
ii) x=0 P(x) = x=0 x! =1
iii) Mean, µ = np = λ
√
iv) Variance, σ 2 = λ, S.D = λ
Relationship to Other Distributions

▶ The Poisson Distribution can be derived as a limiting case of

the Binomial Distribution when the number of trials n
becomes large and the probability of success p becomes small,
such that np = λ remains constant.
▶ For large values of λ, the Poisson Distribution approaches a
Normal Distribution due to the Central Limit Theorem.

Example
If a machine learning model is evaluated over a large dataset with
rare events (such as detecting rare failures), the distribution of the
number of detected failures might be approximated by a Poisson
Distribution.
Applications in AI and Data Science

▶ Event Prediction: The Poisson Distribution is used to

predict the number of events (e.g., clicks, purchases, or
system failures) within a specific time frame.
▶ Anomaly Detection: In network security, the Poisson
Distribution can model the normal number of requests to a
server. Significant deviations might indicate anomalies.
▶ Natural Language Processing (NLP): Used to model the
number of times a rare word occurs in a large text corpus.
▶ Queueing Theory: The Poisson Distribution models the
number of arrivals of jobs or tasks in systems like customer
service queues or server request handling.
Derivation for Mean and Variance of a Poisson Distribution
Let X be a random variable following a Poisson distribution with
parameter λ. The probability mass function (pmf) of X is given by:

e −λ λx
P(X = x) = , x = 0, 1, 2, . . .
x!
1. Mean of a Poisson Distribution
The mean (expected value) E (X ) of a Poisson distribution can be
derived as follows:
∞ ∞
X X e −λ λx
E (X ) = x · P(X = x) = x·
x!
x=0 x=0
λx
Rewriting x as x = x · x(x−1)! , we have:
∞
X λx
E (X ) = e −λ
(x − 1)!
x=1
Let y = x − 1, then x = y + 1 and as x goes from 1 to ∞, y goes
from 0 to ∞:
 ∞
X λy
E (X ) = e −λ λ
y!
y =0
P∞ λy
Using the fact that y =0 y ! = e λ , we get:

E (X ) = e −λ λe λ = λ
Therefore, the mean of a Poisson distribution is:

E (X ) = λ
2. Variance of a Poisson Distribution
The variance Var(X ) of a Poisson distribution is given by
E (X 2 ) − [E (X )]2 .
First, we find E (X 2 ):
∞ ∞
X X e −λ λx
E (X 2 ) = x 2 · P(X = x) = x2 ·
x!
x=0 x=0
Rewriting x 2 as x(x − 1) + x, we have:
∞ ∞
−λ
X λx X λx−1
2
E (X ) = e + e −λ λ
(x − 2)! (x − 1)!
x=2 x=1

Letting y = x − 2 and z = x − 1, we get:

∞ ∞
X λy X λz
E (X 2 ) = e −λ λ2 + e −λ λ
y! z!
y =0 z=0

E (X 2 ) = λ2 + λ
Then, the variance is:

Var(X ) = E (X 2 ) − [E (X )]2 = (λ2 + λ) − λ2 = λ

Thus, the variance of a Poisson distribution is:

Var(X ) = λ
Problem 1

The number of accidents in a year for taxi drivers in a city follows

a Poisson distribution with a mean of 3. Out of 1000 taxi drivers,
find approximately the number of drivers with:
(i) No accident in a year.
(ii) More than 3 accidents in a year.

Solution:
Given that X follows a Poisson distribution with mean λ = 3.

e −λ λx
P(X = x) =
x!
where λ = 3.
(i) Number of drivers with no accident in a year:

e −3 · 30
P(X = 0) = = e −3 ≈ 0.0498
0!
The expected number of drivers with no accidents is:

1000 × P(X = 0) = 1000 × 0.0498 = 49.8 ≈ 50

So, approximately 50 drivers are expected to have no accidents in a
year.

(ii). Number of drivers with more than 3 accidents in a year:

We need to calculate P(X > 3), which is:

P(X > 3) = 1 − P(X ≤ 3)

First, calculate P(X ≤ 3):

P(X ≤ 3) = P(X = 0) + P(X = 1) + P(X = 2) + P(X = 3)

 30
P(X = 0) = e −3 = e −3 ≈ 0.0498
0!
31
P(X = 1) = e −3 = 3e −3 ≈ 0.1494
1!
3 2 9e −3
P(X = 2) = e −3 = ≈ 0.2240
2! 2
33 27e −3
P(X = 3) = e −3 = ≈ 0.2240
3! 6
P(X ≤ 3) ≈ 0.0498 + 0.1494 + 0.2240 + 0.2240 = 0.6472
Thus,

P(X > 3) = 1 − 0.6472 = 0.3528

The expected number of drivers with more than 3 accidents is:

1000 × P(X > 3) = 1000 × 0.3528 = 352.8 ≈ 353

So, approximately 353 drivers are expected to have more than 3
accidents in a year.
Problem 2
In a certain factory turning out razor blades, there is a small
1
probability of 500 for any blade to be defective. The blades are
supplied in packets of 10. Use the Poisson distribution to calculate
the approximate number of packets containing:
(i) No defective blades.
(ii) 2 defective blades.
(iii) 3 defective blades.
in the consignment of 10,000 packets.

Solution:
Let X denote the number of defective blades in a packet. The
1
probability of a blade being defective is p = 500 . The mean
number of defective blades per packet, λ, can be calculated as:
1 10
λ = np = 10 × = = 0.02
500 500
The Poisson probability mass function is given by:

e −λ λx
P(X = x) =
x!
where λ = 0.02.

(i). Number of packets with no defective blades:

e −0.02 · 0.020
P(X = 0) = = e −0.02 ≈ 0.9802
0!
The expected number of packets with no defective blades is:

10000 × P(X = 0) = 10000 × 0.9802 = 9802

So, approximately 9802 packets are expected to have no defective
blades.
(ii). Number of packets with 2 defective blades:

e −0.02 · 0.022
P(X = 2) =
2!
e −0.02 · 0.0004 0.9802 × 0.0004
= ≈ = 0.000196
2 2
The expected number of packets with 2 defective blades is:

10000 × P(X = 2) = 10000 × 0.000196 = 1.96

So, approximately 2 packets are expected to have 2 defective
blades.
(iii). Number of packets with 3 defective blades:

e −0.02 · 0.023
P(X = 3) =
3!
e −0.02 · 0.000008 0.9802 × 0.000008
= ≈ = 0.00000131
6 6
The expected number of packets with 3 defective blades is:

10000 × P(X = 3) = 10000 × 0.00000131 = 0.0131

So, approximately 0 packets are expected to have 3 defective
blades (very unlikely).
Problem 3

If the probability of a bad reaction from a certain injection is

0.001, determine the probability that out of 2000 individuals, more
than 2 will get a bad reaction.

Solution:
Let X denote the number of individuals who get a bad reaction.
The probability of a bad reaction for an individual is p = 0.001.
For 2000 individuals, we have n = 2000.
The mean number of bad reactions, λ, is given by:

λ = np = 2000 × 0.001 = 2
Since n is large and p is small, we can use the Poisson
approximation to the binomial distribution, where X ∼ Poisson(λ).
The probability mass function of a Poisson distribution is:

e −λ λx
P(X = x) =
x!
where λ = 2.
To find the probability that more than 2 individuals will get a bad
reaction, we need to calculate P(X > 2):

P(X > 2) = 1 − P(X ≤ 2)

First, we calculate P(X ≤ 2) using the cumulative distribution
function:

P(X ≤ 2) = P(X = 0) + P(X = 1) + P(X = 2)

Calculating each term:

e −2 · 20
P(X = 0) = = e −2 ≈ 0.1353
0!
 e −2 · 21
P(X = 1) = = 2e −2 ≈ 0.2707
1!

e −2 · 22
P(X = 2) = = 2e −2 ≈ 0.2707
2!
Summing these probabilities:

P(X ≤ 2) = 0.1353 + 0.2707 + 0.2707 = 0.6767

Now, calculate P(X > 2):

P(X > 2) = 1 − 0.6767 = 0.3233

Therefore, the probability that more than 2 individuals will get a
bad reaction is 0.3233.
Assignment Problems
1.The probability that a news reader commits no mistakes in
reading the news is e13 . Find the probability that on a particular
news broadcast he commits:
(i) Only 2 mistakes [Ans: 0.2240]
(ii) More than 3 mistakes [Ans: 0.3500]
(iii) At most 3 mistakes [Ans: 0.6500]
2.Suppose 300 misprints are randomly distributed throughout a
book of 500 pages. Find the probability that a given page contains:
(i) Exactly 3 misprints [Ans: 0.0197]
(ii) Less than 3 misprints [Ans: 0.9768]
(iii) 4 or more misprints [Ans: 0.0033]
3. A certain screw-making machine produces an average of 2
defective screws out of 100 and packs them in boxes of 500. Find
the probability that the box contains:
(i) 3 defective screws [Ans: 0.0075]
(ii) At least 1 defective screw [Ans: 0.9999]
(iii) Between 2 and 4 defective screws [Ans: 0.0287]