PoissonLossFitnessCalculator<T, TInput, TOutput>

Models & Types · AiDotNet.FitnessCalculators

A fitness calculator that uses Poisson Loss to evaluate model performance, particularly for count-based prediction tasks.

Inheritance: FitnessCalculatorBase<T, TInput, TOutput> → PoissonLossFitnessCalculator<T, TInput, TOutput>

For Beginners

This calculator helps evaluate how well your model is performing on count-based prediction tasks, which are problems where you’re predicting the number of times something happens in a fixed interval.

Poisson Loss is designed specifically for problems where:

  • You’re predicting counts or frequencies (whole numbers, 0 or greater)
  • The events occur independently of each other
  • The average rate of events is constant

Examples of count-based prediction problems:

  • Number of customer arrivals per hour
  • Number of goals scored in a soccer match
  • Number of website visits per day
  • Number of defects in a manufacturing process
  • Number of calls to a call center per hour

How Poisson Loss works:

  • It’s based on the Poisson distribution, which models random events occurring over a fixed interval
  • It’s particularly good for data where most values are small counts (0, 1, 2, etc.) but occasionally have larger values
  • It penalizes both overestimation and underestimation, but in a way that’s appropriate for count data

Think of it like this: Imagine you’re predicting how many customers will visit a store each hour:

  • Some hours might have 0 customers
  • Most hours might have 5-10 customers
  • Occasionally, there might be 20+ customers
  • Poisson Loss is designed to handle this kind of pattern well

Key characteristics:

  • It’s specifically designed for count data
  • It works well when the variance of your data increases with the mean
  • Lower values are better (0 would be perfect predictions)
  • It assumes non-negative values (counts can’t be negative)

When to use this calculator:

  • When your target values are counts (whole numbers, 0 or greater)
  • When your data follows a Poisson-like distribution (many small values, fewer large values)
  • When you’re predicting the frequency of events in a fixed time or space interval

Constructors

Constructor Summary
PoissonLossFitnessCalculator(DataSetType dataSetType) Initializes a new instance of the PoissonLossFitnessCalculator class.