Skip to content

Commit 8df5950

Browse files
author
minjk-bl
committed
Fix Evaluation - ROC_Curve, AUC
1 parent 83a17e9 commit 8df5950

2 files changed

Lines changed: 87 additions & 18 deletions

File tree

‎html/m_ml/evaluation.html‎

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -1,20 +1,20 @@
11
<body>
22
<div class="vp-grid-box">
3-
<div class="vp-grid-border-box vp-grid-col-p50">
4-
<label class="vp-orange-text">Import Library</label>
5-
<button class="vp-button vp-import-btn" id="vp_importLibrary">Import</button>
6-
</div>
73
<div class="vp-grid-border-box vp-grid-col-95">
84
<label for="modelType" class="vp-orange-text">Model Type</label>
95
<select id="modelType" class="vp-select vp-state">
106
<option value="rgs">Regression</option>
117
<option value="clf">Classfication</option>
128
<option value="cls">Clustering</option>
139
</select>
14-
<label for="predictData" class="vp-orange-text">Predict Data</label>
15-
<input type="text" class="vp-input vp-state" id="predictData" placeholder="Input predict data" value="pred"/>
1610
<label for="targetData" class="vp-orange-text">Target Data</label>
1711
<input type="text" class="vp-input vp-state" id="targetData" placeholder="Input target data" value="y_test"/>
12+
<label for="predictData" class="vp-orange-text">Predict Data</label>
13+
<input type="text" class="vp-input vp-state" id="predictData" placeholder="Input predict data" value="pred"/>
14+
<label for="model" class="vp-ev-model vp-orange-text">Model</label>
15+
<select id="model" class="vp-ev-model vp-select vp-state">
16+
<!-- Auto create -->
17+
</select>
1818
</div>
1919
<div class="vp-grid-border-box">
2020
<label for="featureData" class="vp-orange-text">Evaluation Metrics</label>
@@ -33,8 +33,8 @@
3333
<label><input type="checkbox" id="precision" class="vp-state"><span>Precision</span></label>
3434
<label><input type="checkbox" id="recall" class="vp-state"><span>Recall</span></label>
3535
<label><input type="checkbox" id="f1_score" class="vp-state"><span>F1-scorev</label>
36-
<label><input type="checkbox" id="roc_curve" class="vp-state"><span>ROC Curve</span></label>
37-
<label><input type="checkbox" id="auc" class="vp-state"><span>AUC</span></label>
36+
<label><input type="checkbox" id="roc_curve" class="vp-eval-check vp-state"><span>ROC Curve</span></label>
37+
<label><input type="checkbox" id="auc" class="vp-eval-check vp-state"><span>AUC</span></label>
3838
</div>
3939
<div class="vp-grid-box vp-eval-box vp-eval-cls">
4040
<label><input type="checkbox" id="silhouetteScore" class="vp-state" checked><span>Silhouette score</span></label>

‎js/m_ml/evaluation.js‎

Lines changed: 79 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -28,20 +28,21 @@ define([
2828
class Evaluation extends PopupComponent {
2929
_init() {
3030
super._init();
31+
this.config.importButton = true;
3132
this.config.dataview = false;
3233

3334
this.state = {
3435
modelType: 'rgs',
3536
predictData: 'pred',
3637
targetData: 'y_test',
38+
// regression
39+
r_squared: true, mae: true, mape: false, rmse: true, scatter_plot: false,
3740
// classification
3841
confusion_matrix: true, report: true,
3942
accuracy: false, precision: false, recall: false, f1_score: false,
40-
// regression
41-
coefficient: false, intercept: false, r_squared: true,
42-
mae: false, mape: false, rmse: true, scatter_plot: false,
43+
roc_curve: false, auc: false,
4344
// clustering
44-
sizeOfClusters: true, silhouetteScore: true,
45+
silhouetteScore: true, ari: false, nm: false,
4546
...this.state
4647
}
4748
}
@@ -63,7 +64,39 @@ define([
6364

6465
$(that.wrapSelector('.vp-eval-box')).hide();
6566
$(that.wrapSelector('.vp-eval-'+modelType)).show();
66-
})
67+
68+
if (modelType == 'clf') {
69+
// Classification - model selection
70+
if (that.checkToShowModel() == true) {
71+
$(that.wrapSelector('.vp-ev-model')).show();
72+
}
73+
}
74+
});
75+
76+
// open model selection show
77+
$(this.wrapSelector('.vp-eval-check')).on('change', function() {
78+
let checked = $(this).prop('checked');
79+
80+
if (checked) {
81+
$(that.wrapSelector('.vp-ev-model')).show();
82+
} else {
83+
if (that.checkToShowModel() == false) {
84+
$(that.wrapSelector('.vp-ev-model')).hide();
85+
}
86+
}
87+
});
88+
}
89+
90+
/**
91+
* Check if anything checked available ( > 0)
92+
* @returns
93+
*/
94+
checkToShowModel() {
95+
let checked = $(this.wrapSelector('.vp-eval-check:checked')).length;
96+
if (checked > 0) {
97+
return true;
98+
}
99+
return false;
67100
}
68101

69102
templateForBody() {
@@ -72,7 +105,7 @@ define([
72105
$(page).find('.vp-eval-box').hide();
73106
$(page).find('.vp-eval-'+this.state.modelType).show();
74107

75-
// varselector TEST:
108+
// varselector
76109
let varSelector = new VarSelector2(this.wrapSelector(), ['DataFrame', 'list', 'str']);
77110
varSelector.setComponentID('predictData');
78111
varSelector.addClass('vp-state vp-input');
@@ -85,6 +118,28 @@ define([
85118
varSelector.setValue(this.state.targetData);
86119
$(page).find('#targetData').replaceWith(varSelector.toTagString());
87120

121+
// model
122+
// set model list
123+
let modelOptionTag = new com_String();
124+
vpKernel.getModelList('Classification').then(function(resultObj) {
125+
let { result } = resultObj;
126+
var modelList = JSON.parse(result);
127+
modelList && modelList.forEach(model => {
128+
let selectFlag = '';
129+
if (model.varName == that.state.model) {
130+
selectFlag = 'selected';
131+
}
132+
modelOptionTag.appendFormatLine('<option value="{0}" data-type="{1}" {2}>{3} ({4})</option>',
133+
model.varName, model.varType, selectFlag, model.varName, model.varType);
134+
});
135+
$(page).find('#model').html(modelOptionTag.toString());
136+
$(that.wrapSelector('#model')).html(modelOptionTag.toString());
137+
138+
if (!that.state.model || that.state.model == '') {
139+
that.state.model = $(that.wrapSelector('#model')).val();
140+
}
141+
});
142+
88143
// load state
89144
let that = this;
90145
Object.keys(this.state).forEach(key => {
@@ -114,8 +169,22 @@ define([
114169
}
115170
});
116171

172+
if (this.state.modelType == 'clf') {
173+
if (this.state.roc_curve == true || this.state.auc == true) {
174+
$(page).find('.vp-ev-model').show();
175+
} else {
176+
$(page).find('.vp-ev-model').hide();
177+
}
178+
} else {
179+
$(page).find('.vp-ev-model').hide();
180+
}
181+
117182
return page;
118183
}
184+
185+
generateImportCode() {
186+
return 'from sklearn import metrics';
187+
}
119188

120189
generateCode() {
121190
let codeCells = [];
@@ -124,6 +193,7 @@ define([
124193
modelType, predictData, targetData,
125194
// classification
126195
confusion_matrix, report, accuracy, precision, recall, f1_score, roc_curve, auc,
196+
model,
127197
// regression
128198
coefficient, intercept, r_squared, mae, mape, rmse, scatter_plot,
129199
// clustering
@@ -173,7 +243,7 @@ define([
173243
if (roc_curve) {
174244
code = new com_String();
175245
code.appendLine("# ROC Curve");
176-
code.appendFormatLine("fpr, tpr, thresholds = roc_curve({0}, svc.decision_function({1}}))", predictData, targetData);
246+
code.appendFormatLine("fpr, tpr, thresholds = metrics.roc_curve({0}, {1}.decision_function({2}))", predictData, model, targetData);
177247
code.appendLine("plt.plot(fpr, tpr, label='ROC Curve')");
178248
code.appendLine("plt.xlabel('Sensitivity') ");
179249
code.append("plt.ylabel('Specificity') ")
@@ -182,8 +252,7 @@ define([
182252
if (auc) {
183253
code = new com_String();
184254
code.appendLine("# AUC");
185-
code.appendFormatLine("fpr, tpr, thresholds = roc_curve({0}, svc.decision_function({1}}))", predictData, targetData);
186-
code.append("metrics.auc(fpr, tpr)");
255+
code.appendFormat("metrics.roc_auc_score({0}, {1}.decision_function({2}))", predictData, model, targetData);
187256
codeCells.push(code.toString());
188257
}
189258
}
@@ -232,7 +301,7 @@ define([
232301
code.appendLine('# Regression plot');
233302
code.appendFormatLine('plt.scatter({0}, {1})', targetData, predictData);
234303
code.appendFormatLine("plt.xlabel('{0}')", targetData);
235-
code.appendFormatLine("plt.ylabel('{1}')", predictData);
304+
code.appendFormatLine("plt.ylabel('{0}')", predictData);
236305
code.append('plt.show()');
237306
codeCells.push(code.toString());
238307
}

0 commit comments

Comments
 (0)