Skip to content

Untitled

Common Definitions

  • \(y\) is the target: No of failures
  • \(x\) is the set of relevant features such as type of transformer, area, temperature

Goal

Obtain a model \(\hat f\) which gives estimates \(\hat y\) from \(x\)

\[ \begin{aligned} \hat y &= \hat f(x) \end{aligned} \]

Poisson is not the most ideal

\[ \begin{aligned} \hat y &= \text{Poisson}(x) \end{aligned} \]
  • Poisson assumes that variance increases with mean
  • Poisson is just an approximation of Binomial distribution when
    • \(n\) is large: many transformers
    • \(p\) is small: low failure rate
  • This approximation is used
    • when \(n\) is unknown, ie no of transformers is unknown
    • for numerical performance, but not a concern nowadays

Better to

\[ \begin{aligned} \hat y &= \hat f(x) \\ &= n \times \hat p(\text{Failure} \vert x_i) \end{aligned} \]
  • Predict the probability of failure of each type of transformer using 'logistic/binomial' model
  • Count of failure = \(n p\)
    • \(n=\) number of transformers
    • \(p =\) probability of failure of each transformer

Advantages - Probability of failure for each \(x\) is insightful - Assumption-free - Better uncertainty quantification

Last Updated: 2024-12-26 ; Contributors: AhmedThahir, web-flow

Comments