Zeroclanzhang（讨论 | 贡献） 无编辑摘要 标签：已被回退 |
Zeroclanzhang（讨论 | 贡献） 无编辑摘要 标签：手工回退 |
||
| 第1行： | 第1行： | ||
{| class="wikitable" align="center" style="border:none; background:transparent; text-align:center;" | <noinclude>{| class="wikitable" align="center" style="border:none; background:transparent; text-align:center;" | ||
| style="border:none;" rowspan="2" | | | style="border:none;" rowspan="2" | | ||
| style="border:none;" | | | style="border:none;" | | ||
| style="background:#bbeeee;" colspan="2" | ''' | | style="background:#bbeeee;" colspan="2" | '''预测条件''' | ||
| style="border:none; text-align:right;" colspan="2" | <sup> | | style="border:none; text-align:right;" colspan="2" | <sup>来源: </sup><ref> | ||
{{cite journal |last=Balayla |first=Jacques |title= | {{cite journal |last=Balayla |first=Jacques |title=流行阈值 (ϕe) 及筛查曲线的几何性质 |journal=PLOS ONE |date=2020 |volume=15 |issue=10 |pages=e0240215 |doi=10.1371/journal.pone.0240215 |pmid=33027310 |doi-access=free }}</ref><ref> | ||
{{cite journal |last=Fawcett |first=Tom |title= | {{cite journal |last=Fawcett |first=Tom |title=ROC 分析简介 |journal=Pattern Recognition Letters |date=2006 |volume=27 |issue=8 |pages=861–874 |doi=10.1016/j.patrec.2005.10.010 |s2cid=2027090 |url=http://people.inf.elte.hu/kiss/11dwhdm/roc.pdf}}</ref><ref> | ||
{{Cite journal|last1=Piryonesi S. Madeh|last2=El-Diraby Tamer E.|date=2020-03-01|title= | {{Cite journal|last1=Piryonesi S. Madeh|last2=El-Diraby Tamer E.|date=2020-03-01|title=资产管理中数据分析：路面状况指数的成本效益预测|journal=Journal of Infrastructure Systems|volume=26|issue=1|pages=04019036|doi=10.1061/(ASCE)IS.1943-555X.0000512|s2cid=213782055 }}</ref><ref> | ||
{{cite journal |first=David M. W. |last=Powers |date=2011 |title= | {{cite journal |first=David M. W. |last=Powers |date=2011 |title=评价：从精确度、召回率和F-度量到ROC、信息度和标记度与相关性 |journal=Journal of Machine Learning Technologies |volume=2 |issue=1 |pages=37–63 |url=https://www.researchgate.net/publication/228529307}}</ref><ref> | ||
{{cite book |last=Ting |first=Kai Ming |editor2-first=Geoffrey I. |editor2-last=Webb |editor1-first=Claude |editor1-last=Sammut |title= | {{cite book |last=Ting |first=Kai Ming |editor2-first=Geoffrey I. |editor2-last=Webb |editor1-first=Claude |editor1-last=Sammut |title=机器学习百科全书 |date=2011 |publisher=Springer |doi=10.1007/978-0-387-30164-8 |isbn=978-0-387-30164-8 }}</ref><ref> | ||
{{cite web |url=https://www.cawcr.gov.au/projects/verification/ |title=WWRP/WGNE | {{cite web |url=https://www.cawcr.gov.au/projects/verification/ |title=WWRP/WGNE 联合预测验证研究工作组 |last1=Brooks |first1=Harold |last2=Brown |first2=Barb |last3=Ebert |first3=Beth |last4=Ferro |first4=Chris |last5=Jolliffe |first5=Ian |last6=Koh |first6=Tieh-Yong |last7=Roebber |first7=Paul |last8=Stephenson |first8=David |date=2015-01-26|website=澳大利亚天气与气候研究合作|publisher=世界气象组织|access-date=2019-07-17}}</ref><ref> | ||
{{cite journal |vauthors = Chicco D, Jurman G |title = | {{cite journal |vauthors = Chicco D, Jurman G |title = 马修斯相关系数 (MCC) 在二元分类评估中优于 F1 分数和准确度的优势 |journal = BMC Genomics |volume = 21 |issue = 1 |date = January 2020 |page = 6-1–6-13 |pmid = 31898477 |doi = 10.1186/s12864-019-6413-7 |pmc = 6941312 |doi-access = free }}</ref><ref> | ||
{{cite journal |vauthors = Chicco D, Toetsch N, Jurman G |title = | {{cite journal |vauthors = Chicco D, Toetsch N, Jurman G |title = 马修斯相关系数 (MCC) 在两类混淆矩阵评估中比平衡精度、博彩信息度和标记度更可靠 |journal = BioData Mining |volume = 14 |issue = 13 |date = February 2021 |page = 13pmid = 33541410 | pmc = 7863449 |doi = 10.1186/s13040-021-00244-z |doi-access = free }}</ref><ref> | ||
{{cite journal |author = Tharwat A. |title = | {{cite journal |author = Tharwat A. |title = 分类评估方法 |journal = Applied Computing and Informatics |date = August 2018 |volume = 17 |pages = 168–192 |doi = 10.1016/j.aci.2018.08.003 |doi-access = free }}</ref> <sup>{{navbar|Diagnostic testing diagram|plain=y}}</sup> | ||
|- | |- | ||
| style="background:#eeeeee;" | [[Statistical population| | | style="background:#eeeeee;" | [[Statistical population|总体人群]] <br/><span style="white-space:nowrap;">= P + N</span> | ||
| style="background:#ccffff;" | ''' | | style="background:#ccffff;" | '''预测阳性 (PP)''' | ||
| style="background:#aadddd;" | ''' | | style="background:#aadddd;" | '''预测阴性 (PN)''' | ||
| style="border-left:double silver;" | [[Youden's_J_statistic| | | style="border-left:double silver;" | [[Youden's_J_statistic|信息度]], {{small|博彩信息度 (BM)}} <br/><span style="white-space:nowrap;">= TPR + TNR − 1</span> | ||
| [[Prevalence threshold]] (PT) <br/><span style="white-space:nowrap;">= | | [[Prevalence threshold]] (PT) <br/><span style="white-space:nowrap;">=[math]\mathsf\tfrac{\sqrt{\text{TPR}\times\text{FPR}}-\text{FPR}}{\text{TPR}-\text{FPR}}[/math]</span> | ||
|- | |- | ||
| rowspan="2" {{verth|va=middle|cellstyle=background:#eeeebb;|''' | | rowspan="2" {{verth|va=middle|cellstyle=background:#eeeebb;|'''实际条件'''}} | ||
| style="background:#ffffcc;" | ''' | | style="background:#ffffcc;" | '''阳性 (P)''' | ||
| style="background:#ccffcc;" | '''[[True positive]] (TP), <br />{{small| | | style="background:#ccffcc;" | '''[[True positive]] (TP), <br />{{small|命中}}''' | ||
| style="background:#ffdddd;" | '''[[False negative]] (FN), <br/>{{small|[[type II error]], | | style="background:#ffdddd;" | '''[[False negative]] (FN), <br/>{{small|[[type II error]], 错失, <br/>低估}}''' | ||
| style="background:#eeffee;" | [[True positive rate]] (TPR), [[recall (information retrieval)| | | style="background:#eeffee;" | [[True positive rate]] (TPR), [[recall (information retrieval)|召回率]], [[Sensitivity (tests)|敏感性]] (SEN), {{small|检测概率, 命中率, [[statistical power|功效]]}} <br/><span style="white-space:nowrap;">= {{sfrac|TP|P}}</span> <span style="white-space:nowrap;">= 1 − FNR</span> | ||
| style="background:#ffeeee;" | [[False negative rate]] (FNR), <br/>{{small| | | style="background:#ffeeee;" | [[False negative rate]] (FNR), <br/>{{small|错失率}} <br/><span style="white-space:nowrap;">= {{sfrac|FN|P}}</span> <span style="white-space:nowrap;">= 1 − TPR</span> | ||
|- | |- | ||
| style="background:#ddddaa;" | ''' | | style="background:#ddddaa;" | '''阴性 (N)''' | ||
| style="background:#ffcccc;" | '''[[False positive]] (FP), <br/>{{small|[[type I error]], | | style="background:#ffcccc;" | '''[[False positive]] (FP), <br/>{{small|[[type I error]], 虚警, <br/>高估}}''' | ||
| style="background:#bbeebb;" | '''[[True negative]] (TN), <br />{{small| | | style="background:#bbeebb;" | '''[[True negative]] (TN), <br />{{small|正确拒绝}}''' | ||
| style="background:#eedddd;" | [[False positive rate]] (FPR), <br/>{{small| | | style="background:#eedddd;" | [[False positive rate]] (FPR), <br/>{{small|虚警概率, [[evaluation measures (information retrieval)#Fall-out|{{nowrap|降雨量}}]]}} <br/><span style="white-space:nowrap;">= {{sfrac|FP|N}}</span> <span style="white-space:nowrap;">= 1 − TNR</span> | ||
| style="background:#ddeedd;"| [[True negative rate]] (TNR), <br/>{{small|[[specificity (tests)| | | style="background:#ddeedd;"| [[True negative rate]] (TNR), <br/>{{small|[[specificity (tests)|特异性]] (SPC), 选择性}} <br/><span style="white-space:nowrap;">= {{sfrac|TN|N}}</span> <span style="white-space:nowrap;">= 1 − FPR</span> | ||
|- | |- | ||
| style="border:none;" rowspan="3"| | | style="border:none;" rowspan="3"| | ||
| style="border-top:double silver; border-right:double silver;"|[[ | | style="border-top:double silver; border-right:double silver;"|[[患病率]] <br/><span style="white-space:nowrap;">= {{sfrac|P|P + N}}</span> | ||
| style="background:#eeffee;" | {{nowrap|[[ | | style="background:#eeffee;" | {{nowrap|[[阳性预测值]] (PPV),}} {{small|[[信息检索中的精确度|精确度]]}} <br/><span style="white-space:nowrap;">= {{sfrac|TP|PP}}</span> <span style="white-space:nowrap;">= 1 − FDR</span> | ||
| style="background:#ffeeee;border-right:double silver;"|[[ | | style="background:#ffeeee;border-right:double silver;"|[[漏报率]] (FOR) <br/><span style="white-space:nowrap;">= {{sfrac|FN|PN}}</span> <span style="white-space:nowrap;">= 1 − NPV</span> | ||
| style="background:#eeeeff;" | [[ | | style="background:#eeeeff;" | [[阳性似然比]] (LR+) <br/><span style="white-space:nowrap;">= {{sfrac|TPR|FPR}}</span> | ||
| style="background:#eeeeff;" | [[ | | style="background:#eeeeff;" | [[阴性似然比]] (LR−) <br/><span style="white-space:nowrap;">= {{sfrac|FNR|TNR}}</span> | ||
|- | |- | ||
| style="border-right:double silver;"|[[ | | style="border-right:double silver;"|[[准确度与精确度#在二元分类中|准确度]] (ACC) <span style="white-space:nowrap;">= {{sfrac|TP + TN|P + N}}</span> | ||
| style="background:#eedddd;"|[[ | | style="background:#eedddd;"|[[假发现率]] (FDR) <br/><span style="white-space:nowrap;">= {{sfrac|FP|PP}}</span> <span style="white-space:nowrap;">= 1 − PPV</span> | ||
| style="background:#ddeedd;"|[[ | | style="background:#ddeedd;"|[[阴性预测值]] (NPV) <span style="white-space:nowrap;">= {{sfrac|TN|PN}}</span> <span style="white-space:nowrap;">= 1 − FOR</span> | ||
| style="border-top:double silver;border-right:double silver;" | [[ | | style="border-top:double silver;border-right:double silver;" | [[标记度]] (MK), {{small|deltaP (Δp)}} <br/><span style="white-space:nowrap;">= PPV + NPV − 1</span> | ||
| style="background:#eeeeff;" | [[ | | style="background:#eeeeff;" | [[诊断比值比|诊断{{nowrap|比值比}}]] (DOR) <span style="white-space:nowrap;">= {{sfrac|LR+|LR−}}</span> | ||
|- | |- | ||
| | | 平衡准确度 (BA) <span style="white-space:nowrap;">= {{sfrac|TPR + TNR|2}}</span> | ||
| style="border-top:double silver;"|[[F1 | | style="border-top:double silver;"|[[F1 分数|F<sub>1</sub> 分数]] <br/><span style="white-space:nowrap;">= {{sfrac|2 PPV × TPR|PPV + TPR}}</span> <span white-space:nowrap;">= {{sfrac|2 TP|2 TP + FP + FN}}</span> | ||
| style="border-top:double silver;"|[[Fowlkes–Mallows | | style="border-top:double silver;"|[[Fowlkes–Mallows 指数]] (FM) <span style="white-space:nowrap;">= [math]\scriptstyle\mathsf\sqrt{\text{PPV}\times\text{TPR}}[/math]</span> | ||
| style="border-top:double silver;"|[[ | | style="border-top:double silver;"|[[马修斯相关系数]] (MCC) <br/>=[math]\scriptstyle\mathsf\sqrt{\text{TPR}\times\text{TNR}\times\text{PPV}\times\text{NPV}}[/math][math]\scriptstyle-\mathsf\sqrt{\text{FNR}\times\text{FPR}\times\text{FOR}\times\text{FDR}}[/math] | ||
| style="border-top:double silver;" colspan="2"| | | style="border-top:double silver;" colspan="2"|威胁分数 (TS), 关键成功指数 (CSI), [[Jaccard 指数#在二元分类混淆矩阵中的Jaccard指数|Jaccard 指数]] <span style="white-space:nowrap;">= {{sfrac|TP|TP + FN + FP}}</span> | ||
|} | |}</noinclude> | ||
</noinclude> | |||
2024年1月25日 (四) 12:51的最新版本
| 预测条件 | 来源: [1][2][3][4][5][6][7][8][9] | ||||
| 总体人群 = P + N |
预测阳性 (PP) | 预测阴性 (PN) | 信息度, 博彩信息度 (BM) = TPR + TNR − 1 |
Prevalence threshold (PT) =[math]\mathsf\tfrac{\sqrt{\text{TPR}\times\text{FPR}}-\text{FPR}}{\text{TPR}-\text{FPR}}[/math] | |
| 阳性 (P) | True positive (TP), 命中 |
False negative (FN), type II error, 错失, 低估 |
True positive rate (TPR), 召回率, 敏感性 (SEN), 检测概率, 命中率, 功效 = TP/P = 1 − FNR |
False negative rate (FNR), 错失率 = FN/P = 1 − TPR | |
| 阴性 (N) | False positive (FP), type I error, 虚警, 高估 |
True negative (TN), 正确拒绝 |
False positive rate (FPR), 虚警概率, 降雨量 = FP/N = 1 − TNR |
True negative rate (TNR), 特异性 (SPC), 选择性 = TN/N = 1 − FPR | |
| 患病率 = P/P + N |
阳性预测值 (PPV), 精确度 = TP/PP = 1 − FDR |
漏报率 (FOR) = FN/PN = 1 − NPV |
阳性似然比 (LR+) = TPR/FPR |
阴性似然比 (LR−) = FNR/TNR | |
| 准确度 (ACC) = TP + TN/P + N | 假发现率 (FDR) = FP/PP = 1 − PPV |
阴性预测值 (NPV) = TN/PN = 1 − FOR | 标记度 (MK), deltaP (Δp) = PPV + NPV − 1 |
诊断比值比 (DOR) = LR+/LR− | |
| 平衡准确度 (BA) = TPR + TNR/2 | F1 分数 = 2 PPV × TPR/PPV + TPR = 2 TP/2 TP + FP + FN |
Fowlkes–Mallows 指数 (FM) = [math]\scriptstyle\mathsf\sqrt{\text{PPV}\times\text{TPR}}[/math] | 马修斯相关系数 (MCC) =[math]\scriptstyle\mathsf\sqrt{\text{TPR}\times\text{TNR}\times\text{PPV}\times\text{NPV}}[/math][math]\scriptstyle-\mathsf\sqrt{\text{FNR}\times\text{FPR}\times\text{FOR}\times\text{FDR}}[/math] |
威胁分数 (TS), 关键成功指数 (CSI), Jaccard 指数 = TP/TP + FN + FP | |
- ↑ Balayla, Jacques (2020). "流行阈值 (ϕe) 及筛查曲线的几何性质". PLOS ONE. 15 (10): e0240215. doi:10.1371/journal.pone.0240215. PMID 33027310.
- ↑ Fawcett, Tom (2006). "ROC 分析简介" (PDF). Pattern Recognition Letters. 27 (8): 861–874. doi:10.1016/j.patrec.2005.10.010. S2CID 2027090.
- ↑ Piryonesi S. Madeh; El-Diraby Tamer E. (2020-03-01). "资产管理中数据分析：路面状况指数的成本效益预测". Journal of Infrastructure Systems. 26 (1): 04019036. doi:10.1061/(ASCE)IS.1943-555X.0000512. S2CID 213782055.
- ↑ Powers, David M. W. (2011). "评价：从精确度、召回率和F-度量到ROC、信息度和标记度与相关性". Journal of Machine Learning Technologies. 2 (1): 37–63.
- ↑ Ting, Kai Ming (2011). Sammut, Claude; Webb, Geoffrey I. (eds.). 机器学习百科全书. Springer. doi:10.1007/978-0-387-30164-8. ISBN 978-0-387-30164-8.
- ↑ Brooks, Harold; Brown, Barb; Ebert, Beth; Ferro, Chris; Jolliffe, Ian; Koh, Tieh-Yong; Roebber, Paul; Stephenson, David (2015-01-26). "WWRP/WGNE 联合预测验证研究工作组". 澳大利亚天气与气候研究合作. 世界气象组织. Retrieved 2019-07-17.
- ↑ Chicco D, Jurman G (January 2020). "马修斯相关系数 (MCC) 在二元分类评估中优于 F1 分数和准确度的优势". BMC Genomics. 21 (1): 6-1–6-13. doi:10.1186/s12864-019-6413-7. PMC 6941312. PMID 31898477.
- ↑ Chicco D, Toetsch N, Jurman G (February 2021). "马修斯相关系数 (MCC) 在两类混淆矩阵评估中比平衡精度、博彩信息度和标记度更可靠". BioData Mining. 14 (13): 13pmid = 33541410. doi:10.1186/s13040-021-00244-z. PMC 7863449.
- ↑ Tharwat A. (August 2018). "分类评估方法". Applied Computing and Informatics. 17: 168–192. doi:10.1016/j.aci.2018.08.003.