🎯 准确率计算器（混淆矩阵与测量误差）Accuracy Calculator (Confusion Matrix & Measurement Error)
两种准确率：分类模型或诊断检测，由 TP、TN、FP、FN 算准确率、精确率、召回率（灵敏度）、特异度、F1 和平衡准确率；测量值，由测量值和真实值算百分误差和准确度。本地处理。Two kinds of accuracy: for a classifier or diagnostic test, accuracy, precision, recall (sensitivity), specificity, F1 and balanced accuracy from TP, TN, FP and FN; for a measurement, the percent error and accuracy from the measured and true values. Processed locally.
使用说明与常见问题How-to & FAQ
「准确率计算器（混淆矩阵与测量误差）」是什么？What is the Accuracy Calculator (Confusion Matrix & Measurement Error) tool?
两种准确率：分类模型或诊断检测，由 TP、TN、FP、FN 算准确率、精确率、召回率（灵敏度）、特异度、F1 和平衡准确率；测量值，由测量值和真实值算百分误差和准确度。本地处理。Two kinds of accuracy: for a classifier or diagnostic test, accuracy, precision, recall (sensitivity), specificity, F1 and balanced accuracy from TP, TN, FP and FN; for a measurement, the percent error and accuracy from the measured and true values. Processed locally.
如何使用「准确率计算器（混淆矩阵与测量误差）」？How do I use the Accuracy Calculator (Confusion Matrix & Measurement Error) tool?
在「准确率计算器（混淆矩阵与测量误差）」中，填写「类型」、「真阳性 TP」、「真阴性 TN」、「假阳性 FP」、「假阴性 FN」等参数，结果会随输入实时计算并显示在下方。In the Accuracy Calculator (Confusion Matrix & Measurement Error): Fill in "Type", "True positives TP", "True negatives TN", "False positives FP", "False negatives FN"; results are recalculated instantly as you type.
准确率高就代表模型好吗？Does high accuracy mean a good model?
不一定。类别很不平衡时（如 95% 是阴性），全猜阴性也有 95% 的准确率。这时要看召回率、精确率、F1 或平衡准确率。Not necessarily. With very unbalanced classes (say 95% negatives), always guessing negative is already 95% accurate. Look at recall, precision, F1 or balanced accuracy instead.
精确率和召回率有什么区别？How do precision and recall differ?
精确率 = 预测为阳性里真正是阳性的比例 TP/(TP+FP)；召回率（灵敏度）= 所有真阳性里被找出来的比例 TP/(TP+FN)。F1 是两者的调和平均。Precision is the share of predicted positives that are really positive, TP/(TP+FP); recall (sensitivity) is the share of actual positives that were found, TP/(TP+FN). F1 is their harmonic mean.
测量的准确度怎么算？How is measurement accuracy calculated?
百分误差 = |测量值 − 真实值| ÷ |真实值| × 100%，准确度 = 100% − 百分误差。例如真实值 10、测量 9.8，误差 2%，准确度 98%。Percent error = |measured − true| ÷ |true| × 100%, and accuracy = 100% − percent error. Measuring 9.8 when the true value is 10 is a 2% error, so 98% accurate.
能举个计算实例吗？Can you show a worked example?
例如在「准确率计算器（混淆矩阵与测量误差）」中填 类型 = 分类 / 检测（混淆矩阵），真阳性 TP = 90，真阴性 TN = 90，假阳性 FP = 10，假阴性 FN = 10，结果区显示：准确率 (TP+TN)/总数：90%；错误率：10%；精确率 TP/(TP+FP)：90%；召回率 / 灵敏度 TP/(TP+FN)：90%；特异度 TN/(TN+FP)：90%；F1 分数：0.9；平衡准确率（灵敏度与特异度的平均）：90%；总数：200For example, entering Type = Classification, True positives TP = 90, True negatives TN = 90, False positives FP = 10, False negatives FN = 10 in the Accuracy Calculator (Confusion Matrix & Measurement Error) gives: Accuracy (TP+TN)/total: 90%; Error rate: 10%; Precision TP/(TP+FP): 90%; Recall / sensitivity TP/(TP+FN): 90%; Specificity TN/(TN+FP): 90%; F1 score: 0.9; Balanced accuracy (mean of sensitivity and specificity): 90%; Total: 200
哪些输入会被检查？Which inputs are validated?
「准确率计算器（混淆矩阵与测量误差）」会检查这些输入——「真阳性 TP」、「真阴性 TN」、「假阳性 FP」、「假阴性 FN」、「真实值（公认值）」：可留空；「测量值」：所选方式用到时必须填写；绝对值超过 10¹⁵ 的数也会被拒绝。不符合时结果区显示具体提示，不会给出数字。The Accuracy Calculator (Confusion Matrix & Measurement Error) checks its inputs — "True positives TP", "True negatives TN", "False positives FP", "False negatives FN", "True (accepted) value": optional; "Measured value": required when the chosen option uses it; numbers above 10¹⁵ in absolute value are rejected too. When a value fails, the result area names the problem instead of showing a number.
有哪些选项可以调整？What options can I adjust?
「准确率计算器（混淆矩阵与测量误差）」的设置：「类型」可选 分类 / 检测（混淆矩阵）、测量值与真实值（默认 分类 / 检测（混淆矩阵））；「真阳性 TP」默认 80；「真阴性 TN」默认 70；「假阳性 FP」默认 10；「假阴性 FN」默认 5；「测量值」默认 9.8；「真实值（公认值）」默认 10。每次生成结果都按当前设置处理。Settings in the Accuracy Calculator (Confusion Matrix & Measurement Error) — "Type": Classification / test (confusion matrix), Measured vs true value (default Classification / test (confusion matrix)); "True positives TP" default 80; "True negatives TN" default 70; "False positives FP" default 10; "False negatives FN" default 5; "Measured value" default 9.8; "True (accepted) value" default 10. The result always uses the current settings.
「准确率计算器（混淆矩阵与测量误差）」免费吗？需要注册吗？Is the Accuracy Calculator (Confusion Matrix & Measurement Error) free? Do I need an account?
「准确率计算器（混淆矩阵与测量误差）」完全免费，无需注册或安装，打开网页即可使用，手机和电脑都能用；输入的内容只在你的浏览器本地处理。The Accuracy Calculator (Confusion Matrix & Measurement Error) is completely free with no sign-up or install and works on phones and desktops; what you enter is processed only in your browser.