demo run also uses parameters from UI where applicable

Author: devhelpr

src/assets/bias-detection-python-code.tsx

@@ -165,58 +165,6 @@ def chi2_test_on_cluster(decoded_X_test, bias_score, cluster_label):
     
     return comparisons
 
-def diffDataframe(df, features, type=None, cluster1=None, cluster2=None):
-    '''
-    Creates difference dataframe, for numerical and categorical 
-    data: Takes dataframe of two clusters of interest and 
-    computes difference in means. Default to analyze most deviating 
-    cluster vs rest of the dataset, except specified otherwise.
-    '''   
-    # Cluster comparison (optional)
-    if cluster1 != None and cluster2 != None:
-        df1 = df[df['Cluster'] == cluster1]
-        df2 = df[df['Cluster'] == cluster2]
-    else:
-        df1 = df[df['Cluster'] == 0]
-        df2 = df[df['Cluster'] != 0]
-
-    n_df1 = df1.shape[0]
-    n_df2 = df2.shape[0]
-
-    diff_dict = {}
-    CI_dict = {}
-
-    for feat in features:
-        sample1 = df1[feat]
-        sample2 = df2[feat]
-
-        if type == 'Numerical':
-            mean1 = np.mean(sample1)
-            mean2 = np.mean(sample2)
-            diff = mean1 - mean2
-            diff_dict[feat] = diff
-        else:
-            freq1 = sample1.value_counts()
-            freq2 = sample2.value_counts()
-            diff = freq1 - freq2
-            diff_dict[feat] = diff
-
-        if type == 'Numerical':
-            pd.set_option('display.float_format', lambda x: '%.5f' % x)
-            diff_df = pd.DataFrame.from_dict(diff_dict, orient='index', columns=['Difference'])
-        else:
-            diff_df = pd.DataFrame()
-            pd.set_option('display.float_format', lambda x: '%.5f' % x)
-
-            for _, value in diff_dict.items():
-                df_temp = pd.DataFrame(value)
-                diff_df = pd.concat([diff_df,df_temp], axis=0,)
-
-            diff_df = diff_df.fillna(0)
-            diff_df.columns = ['Difference']   
-
-    return(diff_df)
-
 def run():
     csv_data = StringIO(data)
     df = pd.read_csv(csv_data)
@@ -227,7 +175,7 @@ def run():
     if isDemo:
         bias_score = "false_positive"
         localDataType = "categorical"
-        localIterations = 20
+        localIterations = iterations # 20
 
         print (f"Using demo parameters: bias_score={bias_score}, targetColumn={targetColumn}, dataType={localDataType}, iterations={iterations}")
 
@@ -297,7 +245,7 @@ def run():
     print(f"X_train shape: {X_train.shape}")
 
     if isDemo:
-        localClusterSize = X_train.shape[0]*0.01
+        localClusterSize = clusterSize # X_train.shape[0]*0.01
     else:
         localClusterSize = clusterSize
 

src/components/BiasSettings.tsx

@@ -8,7 +8,7 @@ import {
 } from '@/components/ui/select';
 import { Slider } from '@/components/ui/slider';
 import { RadioGroup, RadioGroupItem } from '@/components/ui/radio-group';
-import CSVReader, { csvReader } from './CSVReader';
+import CSVReader from './CSVReader';
 import { useEffect, useState } from 'react';
 import { Button } from './ui/button';
 import { ArrowDown, ArrowRight, InfoIcon } from 'lucide-react';
@@ -58,7 +58,14 @@ export default function BiasSettings({
     isErrorDuringAnalysis,
 }: {
     onRun: (params: BiasDetectionParameters) => void;
-    onDataLoad: csvReader['onChange'];
+    onDataLoad: (
+        data: Record<string, string>[],
+        stringified: string,
+        fileName: string,
+        demo?: boolean,
+        columnsCount?: number,
+        params?: BiasDetectionParameters
+    ) => void;
     isLoading: boolean;
     isErrorDuringAnalysis: boolean;
     isInitialised: boolean;
@@ -144,7 +151,18 @@ export default function BiasSettings({
             file.data as Record<string, string>[],
             Papa.unparse(file.data),
             'demo',
-            true
+            true,
+            undefined,
+            {
+                clusterSize: clusters[0],
+                iterations: iter[0],
+                targetColumn: '',
+                dataType: '',
+                higherIsBetter:
+                    form.getValues().whichPerformanceMetricValueIsBetter ===
+                    'higher',
+                isDemo: true,
+            }
         );
     };
 

src/routes/BiasDetection.tsx

@@ -2,7 +2,6 @@ import { useEffect, useRef, useState } from 'react';
 import { pythonCode } from '@/assets/bias-detection-python-code';
 import { usePython } from '@/components/pyodide/use-python';
 import BiasSettings from '@/components/BiasSettings';
-import { csvReader } from '@/components/CSVReader';
 import { cn } from '@/lib/utils';
 import ComponentMapper from '@/components/componentMapper';
 import { useReactToPrint } from 'react-to-print';
@@ -45,11 +44,14 @@ const PAGE_STYLE = `
 `;
 
 export default function BiasDetection() {
-    const [data, setData] = useState<CSVData>({
+    const [data, setData] = useState<
+        CSVData & { params?: BiasDetectionParameters }
+    >({
         data: [],
         stringified: '',
         fileName: '',
         demo: false,
+        params: undefined,
     });
     const { t, i18n } = useTranslation();
 
@@ -80,14 +82,15 @@ export default function BiasDetection() {
         higherIsBetter: false,
         isDemo: false,
     });
-
-    const onFileLoad: csvReader['onChange'] = (
-        data,
-        stringified,
-        fileName,
-        demo
-    ) => {
-        setData({ data, stringified, fileName, demo });
+    const onFileLoad: (
+        data: Record<string, string>[],
+        stringified: string,
+        fileName: string,
+        demo?: boolean,
+        columnsCount?: number,
+        params?: BiasDetectionParameters
+    ) => void = (data, stringified, fileName, demo, _columnsCount, params) => {
+        setData({ data, stringified, fileName, demo, params });
     };
 
     useEffect(() => {
@@ -108,15 +111,8 @@ export default function BiasDetection() {
         if (pythonCode && data.stringified.length >= 0 && initialised) {
             sendData(data.stringified);
         }
-        if (data.demo) {
-            onRun({
-                iterations: 3,
-                clusterSize: 3,
-                targetColumn: 'FP',
-                dataType: 'numeric',
-                higherIsBetter: true,
-                isDemo: true,
-            });
+        if (data.demo && data.params) {
+            onRun(data.params);
         }
     }, [initialised, data]);