Update my_pages/multiverse.py

my_pages/multiverse.py

@@ -249,31 +249,45 @@ def render():
     
         return split_and_scale(features, label, test_split, preprocess_scale)
     
+    def model_train_and_pred(scaler, feature_sel, arch, seed):
+        X_train, X_test, y_train, y_test = get_stackoverflow_dataset(preprocess_scale=scaler)
+
+        if feature_sel=="Select K Best (k=5)":
+            selector = SelectKBest(score_func=f_classif, k=5)
+            X_train = selector.fit_transform(X_train, y_train)
+            X_test = selector.transform(X_test)
+        elif feature_sel=="PCA (n=5)":
+            pca = PCA(n_components=2)
+            X_train = pca.fit_transform(X_train, y_train)
+            X_test = pca.transform(X_test)
+        
+        modelclass_dict = {'Neural Network (Small)': MLPClassifier([10], random_state=seed, max_iter=500),
+                           'Logistic Regression': SGDClassifier(random_state=seed, max_iter=500),
+                           'Decision Tree': DecisionTreeClassifier(random_state=seed)}
+        model = modelclass_dict[arch]
+        model.fit(X_train, y_train)
     
+        y_pred = model.predict(X_test)
+        return y_pred
+        
     ### Main Code Starts Here
     scaler, feature_sel, arch, seed = selected_path[0], selected_path[1], selected_path[2], int(selected_path[3])
-    X_train, X_test, y_train, y_test = get_stackoverflow_dataset(preprocess_scale=scaler)
-
-    if feature_sel=="Select K Best (k=5)":
-        selector = SelectKBest(score_func=f_classif, k=5)
-        X_train = selector.fit_transform(X_train, y_train)
-        X_test = selector.transform(X_test)
-    elif feature_sel=="PCA (n=5)":
-        pca = PCA(n_components=2)
-        X_train = pca.fit_transform(X_train, y_train)
-        X_test = pca.transform(X_test)
-    
-    placeholder = st.empty()
-    modelclass_dict = {'Neural Network (Small)': MLPClassifier([10], random_state=seed, max_iter=500),
-                       'Logistic Regression': SGDClassifier(random_state=seed, max_iter=500),
-                       'Decision Tree': DecisionTreeClassifier(random_state=seed)}
-    model = modelclass_dict[arch]
-    placeholder.write("Training your model.")
-    model.fit(X_train, y_train)
-    placeholder.empty()
-
-    y_pred = model.predict(X_test)
 
+    all_preds = []
+    for scaler in choices_list[0]["options"]:
+        for feature_sel in choices_list[1]["options"]:
+            for arch in choices_list[2]["options"]:
+                for seed in choices_list[3]["options"]:
+                    seed = int(seed)
+                    y_pred_local = model_train_and_pred(scaler, feature_sel, arch, seed)
+                    all_preds.append(y_pred_local)
+                st.markdown(scaler + feature_sel + arch)
+                all_preds_numpy = np.array(all_preds)
+                np.save("all_predictions.npy", all_preds_numpy)
+
+    all_preds_numpy = np.array(all_preds)
+    np.save("all_predictions.npy", all_preds_numpy)
+    
     uniq_perc = 10.5
     add_red_text(f"""
         <b>Multiplicity of your predictions:</b> Your chosen model rejected {uniq_perc}% loans that would have been accepted by more than 90% of the other models.<br><br>