Accuracy refers to the proportion of correctly classified samples in all classifications. For example, if a classification model includes samples of Class A and Class B, and the model correctly identifies A0 samples of Class A and B0 samples of Class B, then the accuracy is (A0+B0)/(A+B)? .
Extended data:
After the initial retrieval is completed, it is necessary to evaluate the recall rate of the initial retrieval of the patent literature database, which is the basis for indicating whether the retrieval can be completed. If the recall rate is not ideal at this time, it is necessary to continue the recall work, otherwise, if the expected recall rate is reached, the recall work can be ended.
Denoising process, also known as "accurate" process, is the process of removing patent documents irrelevant to the analysis topic from the search database. In this process, the effective documents will inevitably be deleted by mistake. In order to check whether too many valid documents were deleted by mistake during denoising, it is necessary to evaluate the recall rate of patent documents after denoising at the end of the work.