31:T52d,view = Plot.plot({ title: "Distribution of Scores", fontFamily: "Roboto", style: {fontSize: "14px"}, height: 600, width: 1080, x: {domain: [0, 100]}, color: { range: ["red", "red", "green", "green"], legend: false, columns: 2, }, symbol: { legend: false, range: ["circle", "square"] }, marks: [ Plot.dot( vizData.join_full(fullType, "type"), Plot.dodgeY({ x: "score", fill: "type", fillOpacity: 0.55, anchor: "bottom", r: radius, title: d => `${d.type}`, stroke: "black", strokeWidth: 1, symbol: "class", }) ), Plot.ruleX([range], {strokeWidth: 2}), // Plot.text( // values[0], // {x: range, y: 10, text: "Hello"} // ), Plot.area( {x1: 0, x2: range, y1: 0, y2: 100, fill: "red", opacity: 0.5, } ), Plot.text( vizData.join_full(fullType, "type"), Plot.dodgeY({ x: "score", fill: "type", r: radius, anchor: "bottom", text: "type", stroke: "black", fill: "black", strokeWidth: 0.1, fontSize: radius - 1, pointerEvents: "none" }) ), Plot.axisX({anchor: "top", labelOffset: -10}), ], tooltip: { opacity: 0.5, stroke: "transparent" }, })32:T598,confusionMatrix = htl.html`

		Actual Class
		positive	negative
Predicted Class	positive	TP: ${TP}	FP: ${FP}	Precision = ${Math.round(TP/(TP+FP) * 100) + "%"}
Predicted Class	negative	FN: ${FN}	TN: ${TN}	NPV = ${Math.round(TN/(TN+FN)*100) + "%"}
		Sensitivity = ${Math.round(TP/(TP+FN)*100) + "%"}	Specificity = ${Math.round(TN/(TN+FP)*100) + "%"}	Accuracy = ${Math.round((TN + TP)/(TN+FP+FN+TP)*100) + "%"}

${ROC}

${PR}

`33:T69d,CutOffData = aq.from(data) .derive({ score: aq.escape(d => radios === "Example 1" ? d.target_low_score * 100 : radios === "Example 2" ? d.target_high_score * 100 : radios === "Example 3" ? d.target_low_score_bad * 100 : d.target_high_score_bad * 100 ), class: aq.escape(d => radios === "Example 1" ? d.target_low : radios === "Example 2" ? d.target_high : radios === "Example 3" ? d.target_low : d.target_high ), }) .derive({ identity: aq.escape(d => "1"), }) .join_full(cutOffs, "identity") .derive({ classPredicted: aq.escape(d => d.score > d.cutOff ? "successful" : "unsuccessful"), }) .derive({ type: d => d.class === "successful" && d.classPredicted === "successful" ? "TP" : d.class === "unsuccessful" && d.classPredicted === "successful" ? "FP" : d.class === "unsuccessful" && d.classPredicted === "unsuccessful" ? "TN" : d.class === "successful" && d.classPredicted === "unsuccessful" ? "FN": null, }) .groupby(["type", "cutOff"]) .rollup({ count: d => op.count(), }) .join_full(fullType, "type") .derive({ count: d => d.count === undefined ? 0 : d.count, }) .orderby(["type", "cutOff"]) .groupby("cutOff") .pivot("type", "count" ) // { count: d => d.count === undefined ? 0 : d.count} .derive({ TN: d => d.TN === undefined ? 0 : d.TN, TP: d => d.TP === undefined ? 0 : d.TP, FN: d => d.FN === undefined ? 0 : d.FN, FP: d => d.FP === undefined ? 0 : d.FP, }) .derive({ Sensitivity: d => d.TP/(d.TP+d.FN)*100, "1-Specificity": d => 100 - d.TN/(d.TN+d.FP)*100, Precision: d => d.TP/(d.TP+d.FP)*100, })34:T551,html`

		Tatsächliche Klasse
		positive	negative
Prognostizierte Klasse	positive	# richtig positive (TP)	# falsch positive (FP)	${tex`Precision = \frac{TP}{TP+FP}`}
Prognostizierte Klasse	negative	# falsch negative (FN)	# richtig negative (TN)	${tex`NPV = \frac{TN}{TN+FN}`}
		${tex`Sensitivity = \frac{TP}{TP+FN}`}	${tex`Specificity = \frac{TN}{TN+FP}`}	${tex`Accuracy = \frac{TP+TN}{n}`}

`35:[{"id":"a310dedc1694c88d065640835f562de5e02afb62a6c9c5e03e6fa9eacafaedd15f421fc89e66715338a886750202685434dda8174934cc3ab8f65609dcc8464b","href":"https://static.observableusercontent.com/files/a310dedc1694c88d065640835f562de5e02afb62a6c9c5e03e6fa9eacafaedd15f421fc89e66715338a886750202685434dda8174934cc3ab8f65609dcc8464b","name":"cutOffData.csv","type":"text/csv","lastModified":1708527269919.221,"size":42814}] 2f:{"title":"Cut-Off-Point, Confusion Matrix, ROC- and PR-Curve","cells":[{"id":0,"value":"# Cut-Off-Point, Confusion Matrix, ROC- and PR-Curve","pinned":false,"hidden":false,"mode":"md"},{"id":164,"value":"$31","pinned":false,"hidden":false,"mode":"ojs"},{"id":195,"value":"viewof range = Inputs.range([-1, 101], {label: null, step: 1, width: 961, value: 50, format: x => Math.round(x)})","pinned":false,"hidden":false,"mode":"ojs"},{"id":177,"value":"viewof radios = Inputs.radio([\"Example 1\", \"Example 2\", \"Example 3\", \"Example 4\"], {label: null, value: \"Example 1\"})","pinned":false,"hidden":false,"mode":"ojs"},{"id":1094,"value":"$32","pinned":false,"hidden":false,"mode":"ojs"},{"id":356,"value":"---\n\n\n# Appendix\n","pinned":false,"hidden":false,"mode":"md"},{"id":365,"value":"\n\n","pinned":false,"hidden":false,"mode":"html"},{"id":713,"value":"ROC = Plot.plot({\n title: \"Receiver Operating Characteristic\",\n fontFamily: \"Roboto\",\n style: {fontSize: \"14px\"},\n width: 300,\n height: 300,\n marks: [\n Plot.ruleY([0]),\n Plot.lineY(CutOffData, {x: \"1-Specificity\", y: \"Sensitivity\" , }),\n // Plot.areaY(CutOffData.orderby(\"Sensitivity\"), {y: \"Sensitivity\", fillOpacity: 0.3}),\n Plot.dot([1], {x: (100 - TN/(TN+FP) * 100), y: TP/(TP+FN)*100, fill: \"blue\", r: 5, }),\n Plot.line([{x: 0, y: 0}, {x: 100, y: 100}], {stroke: \"black\", x: \"x\", y: \"y\", strokeDasharray: 4}),\n Plot.axisX({labelOffset: -10, label: \"1-Specificity\"}),\n Plot.axisY({labelOffset: -10, label: \"Sensitivity\"})\n ]\n})","pinned":false,"hidden":false,"mode":"ojs"},{"id":721,"value":"PR = Plot.plot({\n title: \"Precision-Recall\",\n fontFamily: \"Roboto\",\n style: {fontSize: \"14px\"},\n width: 300,\n height: 300,\n marks: [\n Plot.ruleY([0]),\n Plot.lineY(CutOffData, {y: \"Precision\", x: \"Sensitivity\"}),\n Plot.dot([1], {y: TP/(TP+FP) * 100, x: TP/(TP+FN)*100, fill: \"blue\", r: 5, }),\n Plot.line([{x: 0, y: noInformationRate}, {x: 100, y: noInformationRate}], {stroke: \"black\", x: \"x\", y: \"y\", strokeDasharray: 4}),\n Plot.axisX({labelOffset: -10, label: \"Sensitivity (Recall)\"}),\n Plot.axisY({labelOffset: -10, label: \"Precision\"})\n ]\n})","pinned":false,"hidden":false,"mode":"ojs"},{"id":847,"value":"noInformationRate = Math.min(\n ... vizData\n .groupby(\"class\")\n .rollup({\n noInformationRate: d => op.count()\n })\n .objects()\n .map(d => d.noInformationRate)\n ) /\n numberOfObservations * 100 ","pinned":false,"hidden":false,"mode":"ojs"},{"id":874,"value":"numberOfObservations = vizData.rollup({n: d => op.count()}).objects().map(d => d.n)[0]","pinned":false,"hidden":false,"mode":"ojs"},{"id":656,"value":"$33","pinned":false,"hidden":false,"mode":"ojs"},{"id":643,"value":"cutOffs = aq.table({ cutOff: Array.from(Array(101).keys()) }) \n .derive({\n identity: aq.escape(d => \"1\"),\n })","pinned":false,"hidden":false,"mode":"ojs"},{"id":202,"value":"values = vizData\n .groupby(\"type\")\n .rollup({\n count: d => op.count(),\n })\n .join_full(fullType, \"type\")\n .derive({\n count: d => d.count === undefined ? 0 : d.count,\n })\n .orderby(\"type\")\n .objects()","pinned":false,"hidden":false,"mode":"ojs"},{"id":296,"value":"TN = values.filter(d => d.type === \"TN\").map(d => d.count)[0]","pinned":false,"hidden":false,"mode":"ojs"},{"id":298,"value":"FN = values.filter(d => d.type === \"FN\").map(d => d.count)[0]","pinned":false,"hidden":false,"mode":"ojs"},{"id":287,"value":"TP = values.filter(d => d.type === \"TP\").map(d => d.count)[0]","pinned":false,"hidden":false,"mode":"ojs"},{"id":293,"value":"FP = values.filter(d => d.type === \"FP\").map(d => d.count)[0]","pinned":false,"hidden":false,"mode":"ojs"},{"id":572,"value":"radius = 9.5","pinned":false,"hidden":false,"mode":"ojs"},{"id":268,"value":"fullType = aq.from([\n {type: \"TP\"},\n {type: \"FP\"},\n {type: \"FN\"},\n {type: \"TN\"},\n])","pinned":false,"hidden":false,"mode":"ojs"},{"id":182,"value":"vizData = aq.from(data)\n .derive({\n score: aq.escape(d => \n radios === \"Example 1\" ? d.target_low_score * 100 : \n radios === \"Example 2\" ? d.target_high_score * 100 :\n radios === \"Example 3\" ? d.target_low_score_bad * 100 :\n d.target_high_score_bad * 100\n ),\n class: aq.escape(d => \n radios === \"Example 1\" ? d.target_low : \n radios === \"Example 2\" ? d.target_high :\n radios === \"Example 3\" ? d.target_low :\n d.target_high\n ),\n })\n .derive({\n classPredicted: aq.escape(d => d.score > range ? \"successful\" : \"unsuccessful\"),\n })\n .derive({\n type: d => \n d.class === \"successful\" && d.classPredicted === \"successful\" ? \"TP\" :\n d.class === \"unsuccessful\" && d.classPredicted === \"successful\" ? \"FP\" :\n d.class === \"unsuccessful\" && d.classPredicted === \"unsuccessful\" ? \"TN\" :\n d.class === \"successful\" && d.classPredicted === \"unsuccessful\" ? \"FN\": null,\n })","pinned":false,"hidden":false,"mode":"ojs"},{"id":166,"value":"data = FileAttachment(\"cutOffData.csv\").csv({ typed: true })","pinned":false,"hidden":false,"mode":"ojs"},{"id":184,"value":"import {aq, op} from '@uwdata/arquero'","pinned":false,"hidden":false,"mode":"ojs"},{"id":254,"value":"import {Plot} from \"@mkfreeman/plot-tooltip\"","pinned":false,"hidden":false,"mode":"ojs"},{"id":11,"value":"$34","pinned":false,"hidden":false,"mode":"ojs"}],"files":"$W35","stdlib":"1","titling":"auto"}