Precision ratio (P for short) refers to the percentage of detected related documents to the total number of detected documents. The precision rate reflects the accuracy of retrieval, and its supplement is the false detection rate. Recall rate (R for short) refers to the percentage of related documents detected in the system. The recall rate reflects the comprehensiveness of retrieval, and its supplement is the missed detection rate.
Recall ratio = (the amount of relevant information retrieved/the total amount of relevant information in the system) * 100%. Accuracy = (relevant information retrieved/total information retrieved) * 100%. Using these two quantitative indicators, the performance level of information retrieval system can also be evaluated. In order to evaluate the performance level of an information retrieval system, it is necessary to conduct multiple searches in a retrieval system.
For each search, the precision and recall are calculated as coordinate values and marked on the plane coordinate map. Through a large number of retrieval, the performance curve of the retrieval system can be obtained.
Experiments show that there is an opposite interdependence between recall and accuracy-if the output recall increases, its accuracy will decrease, and vice versa. Understand the relationship between recall rate and accuracy. If the recall rate increases, it means that the shadow part in the middle becomes larger, and the total amount of related documents in the system should remain unchanged. However, the improvement of accuracy is related to the total number of tests. In fact, if you want to find more relevance, there will be more irrelevant detection, that is, the light blue part of the picture will also become larger, resulting in lower accuracy.