混淆矩阵的查准率和查全率

Question

Assume you are solving a 4-class problem. Your test set is as follows: 
•   5 samples from class 1, 
•   10 samples from class 2, 
•   5 samples from class 3,
•   10 samples from class 4. 
•   Total Samples: 30
The decision made by your classifier is as follows:
•   2 samples from class 1 are decided as class 1, 3 samples from class 1 are decided as class 2.
•   2 samples from class 2 are decided as class 1, 5 samples from class 2 are decided as class 2, 1 sample from class 2 are decided as class 3, and 2 samples from class 2 is decided as class 4.
•   4 samples from class 3 are decided as class 3 and 1 sample from class 3 is decided as class 4.
•   2 samples from class 4 are decided as class 1, and 8 samples from class 4 are decided as class 4.
Generate a confusion matrix. Using the confusion matrix, calculate accuracy, average precision, and average recall rate.

我需要帮助使用下面的混淆矩阵手工计算准确度、召回率和精度。

Answer 1

精度:对角线上的数字之和，除以网格上所有数字的和。

召回和精确取决于你是想采取微观还是宏观的方法。有关更多细节，请参见这篇博客文章(它给出了一个与您的案例非常相似的示例)：https://towardsdatascience.com/confusion-matrix-for-your-multi-class-machine-learning-model-ff9aa3bf7826