Precision
The share of actually positive cases among all cases the model marks as positive.
How reliable is the positive label?
Precision measures what share of positively labelled cases are actually positive according to the defined reference. The formula is TP / (TP + FP): TP means true-positive and FP false-positive assignments. The metric belongs to the selected class and assessment. If no case was labelled positive, the denominator is zero; reporting must make the handling of such an undefined value clear.
In a fictional calculation with 50 service requests, a model marks 24 as belonging to technical service. Under the previously agreed assignment rule, 18 actually belong there and 6 are incorrectly marked. Precision is 18 / 24 = 75 %. For operations, this view makes the additional misrouted cases visible. The metric alone does not show whether technical requests were also missed; recall is examined for that purpose.
Sources
Related terms
- LLM evaluationLLM evaluation tests model outputs against defined tasks and criteria. One good example does not establish overall quality.
- Test data in machine learningData reserved for final assessment whose results were not used to select the model variant being assessed.
- RecallThe share of actually positive cases that the model correctly identifies as positive.
- F1 scoreThe harmonic mean of precision and recall for the same positive class.
- Confusion matrixA table comparing reference classes with model predictions and counting correct and incorrect assignments.
- Re-rankingRe-ranking reassesses search candidates already retrieved. Relevance to the query and the limits of the initial selection both matter.
Your project with ASES
Describe your project to us. We will get back to you with an initial assessment.