Poisson Distribution
What you'll learn: How to model the probability of rare, independent events occurring within a fixed interval using the Poisson distribution.
The Core Idea
Imagine you're tracking how many customers enter a quiet coffee shop per hour, or how many typos appear per page in a manuscript. These are count events that happen randomly but with some average rate. The Poisson distribution helps us predict the probability of seeing exactly k events in a fixed window when we know the average rate λ (lambda).
Unlike the Binomial distribution (which counts successes in n trials), Poisson models situations where:
- Events are rare relative to possible opportunities
- Events occur independently
- The average rate stays constant
The Formula
The probability of observing exactly k events is:
P(X = k) = (λ^k × e^(-λ)) / k!
Where:
- λ (lambda) = average number of events in the interval
- k = the specific count you're asking about (0, 1, 2, ...)
- e ≈ 2.71828 (the mathematical constant)
Real-World Example
If a website averages λ = 3 crashes per month, the Poisson distribution tells you:
- P(0 crashes) = probability of a perfect month
- P(5 crashes) = probability of a rough month
- The expected value (mean) = λ = 3
- The variance also = λ = 3
Why It Matters in ML
Poisson distributions model count data in recommendation systems (clicks per session), anomaly detection (rare failure events), and natural language processing (word occurrences in documents).
Key Takeaway: The Poisson distribution models how many times rare, independent events occur in a fixed interval, characterized entirely by its average rate λ.