Recently, I have been doing research and application work on related recommendations. The concepts of recall rate and accuracy rate are occasionally encountered, and I know the meaning, but sometimes it is a bit difficult to introduce them to my classmates clearly.
Recall rate and accuracy rate are two concepts and indicators that are often involved in prediction in data mining and search engines in the Internet.
Recall rate : Recall, also known as "recall rate" - or recall rate is easy to remember, and it can better reflect its substantial meaning.
Accuracy : Precision , also known as "precision", "correct rate".
Taking retrieval as an example, the search situation can be represented by the following figure:
|
related
|
irrelevant
|
|
|
retrieved
|
A
|
B
|
|
not retrieved
|
C
|
D
|
A: Retrieved, relevant (searched and wanted) B: Retrieved, but irrelevant (searched but useless) C: Not retrieved, but relevant (nothing Searched , but actually wanted) D: Not retrieved, not relevant (not found and useless) If we hope: the more content is retrieved, the better, this is the pursuit of "find all" rate", that is, A /(A+C), the bigger the better. If we hope that in the retrieved documents, the more relevant, the better, the less irrelevant the better, this is the pursuit of "accuracy", that is, A/(A+B), the more Bigger is better. Although there is no necessary relationship between "recall rate" and "accuracy rate" (as can be seen from the above formula), in practical applications, they are mutually restrictive . According to actual needs, find a balance point. Often difficult to react quickly is "recall". I guess it also has something to do with the literal meaning, the meaning of which cannot be directly seen from the literal meaning . "Recall" in Chinese means: to bring xx back. The English "recall" corresponding to " recall rate", recall has the meaning of "remember" in addition to the meaning of "order sth to return" mentioned above. Recall: the ability to remember sth. that you have learned or sth. that has happened in the past. When we ask the retrieval system for all the details of a certain thing (enter the retrieval query), Recall means: the retrieval system can "
How many details of those events can be recalled, colloquially, is the “ability to recall”. Dividing the “number of details that can be recalled” by “the system knows all the details of the event” is the “memory rate”, or recall— —Recall rate. Simply, it can also be understood as recall rate.