added weights, converted data to arrow friendly format

utils.py

@@ -35,7 +35,8 @@ def transformation(input, categories):
     match_index = np.where(categories == cat)[0]
     result_array[match_index] = 1
     new_x.extend(result_array.tolist())
-    return new_x
+    python_objects = [np_type.item() if isinstance(np_type, np.generic) else np_type for np_type in new_x]
+    return python_objects
 
 def get_request_body(datapoint):
     data = datapoint.iloc[0].tolist()
@@ -51,8 +52,10 @@ def get_explainability_texts(shap_values, feature_texts):
     sorted_positive_indices = [index for index, _ in sorted(positive_dict.items(), key=lambda item: abs(item[1]), reverse=True)]
     positive_texts = [feature_texts[x] for x in sorted_positive_indices]
     positive_texts = positive_texts[2:]
+    sorted_positive_indices = sorted_positive_indices[2:]
     if len(positive_texts) > 5:
         positive_texts = positive_texts[:5]
+        sorted_positive_indices = sorted_positive_indices[:5]
     return positive_texts, sorted_positive_indices
 
 
@@ -67,11 +70,21 @@ def get_explainability_values(pos_indices, datapoint):
         else:
             val = transformed_data[idx]
         vals.append(val)
-    vals = vals[2:]
-    if len(vals) > 5:
-        vals = vals[:5]
     return vals
 
+# def get_weights(shap_values, sorted_indices):
+#     weights = [shap_values[x] for x in sorted_indices]
+#     total_sum = sum(weights)
+#     scaled_values = [val/total_sum for val in weights]
+#     return scaled_values
+
+def get_weights(shap_values, sorted_indices, target_sum=0.95):
+    weights = [shap_values[x] for x in sorted_indices]
+    total_sum = sum(weights)
+    # Scale to the target sum (0.95 in this case)
+    scaled_values = [val * (target_sum / total_sum) for val in weights]
+    return scaled_values
+
 def get_fake_certainty():
     # Generate a random certainty between 75% and 99%
     fake_certainty = uniform(0.75, 0.99)
@@ -129,18 +142,26 @@ def get_comment_explanation(certainty, explainability_texts, explainability_valu
     return comment
 
 def create_data_input_table(datapoint, col_names):
-    st.subheader("Flagged Transaction:")
+    st.subheader("Transaction details")
     data = datapoint.iloc[0].tolist()
     data[7:12] = [bool(value) for value in data[7:12]]
     rounded_list = [round(value, 2) if isinstance(value, float) else value for value in data]
     df = pd.DataFrame({"Feature name": col_names, "Value": rounded_list })
-    st.dataframe(df, hide_index=True, width=450, height=35*len(df)+38)
+    st.dataframe(df, hide_index=True, use_container_width=True, height=35*len(df)+38) #width=450
 
 # Create a function to generate a table
-def create_table(texts, values, title):
-    df = pd.DataFrame({"Feature Explanation": texts, 'Value': values})
+def create_table(texts, values, weights, title):
+    df = pd.DataFrame({"Feature Explanation": texts, 'Value': values, 'Weight': weights})
     st.markdown(f'#### {title}')  # Markdown for styling
-    st.dataframe(df, hide_index=True, width=450)  # Display a simple table
+    st.dataframe(df, hide_index=True, use_container_width=True, column_config={
+                     'Weight': st.column_config.ProgressColumn(
+                         'Weight',
+                         width='small',
+                         format="%.2f",
+                         min_value=0,
+                         max_value=1
+                     )
+                 }) #width=450  # Display a simple table
 
 
 def ChangeButtonColour(widget_label, font_color, background_color='transparent'):