Update app.py

app.py

@@ -1,22 +1,27 @@
+# %%
 import gradio as gr
-import tensorflow as tf
+from sklearn.ensemble import RandomForestRegressor
 import numpy as np
 import pandas as pd
-
 import pickle
 
-# Define the filename of your pickle model
+# %%
+# TODO change the file to your own model.
 model_filename = "random_forest_regression_luxurious.pkl"
 
-# Open the file in read-binary mode and load the model
+random_forest_model = RandomForestRegressor()
 with open(model_filename, 'rb') as f:
-    model = pickle.load(f)
+    random_forest_model = pickle.load(f)
 
-print("Pickle model loaded successfully!")
+print('Number of features: ', random_forest_model.n_features_in_)
+print('Features are (see week 1): ', ['rooms', 'area', 'pop', 'pop_dens', 'frg_pct', 'emp', 'tax_income', 'luxurious'])
+random_forest_model
 
+# %%
 df_bfs_data = pd.read_csv('bfs_municipality_and_tax_data.csv', sep=',', encoding='utf-8')
 df_bfs_data['tax_income'] = df_bfs_data['tax_income'].str.replace("'", "").astype(float)
 
+# %%
 locations = {
     "Zürich": 261,
     "Kloten": 62,
@@ -124,18 +129,21 @@ locations = {
     "Horgen": 295
 }
 
-# Update the prediction function to include the "luxurious" input
+# %%
+# Define the core prediction function including the "luxurious" input
 def predict_apartment(rooms, area, town, luxurious):
     bfs_number = locations[town]
-    df = df_bfs_data[df_bfs_data['bfs_number'] == bfs_number]
-    
+    df = df_bfs_data[df_bfs_data['bfs_number'] == bfs_number].copy()
+    df.reset_index(inplace=True)
+    df.loc[0, 'rooms'] = rooms
+    df.loc[0, 'area'] = area
     if len(df) != 1:  # if not exactly one record, return -1
         return -1
-    
+
     # Convert the luxurious input (a boolean from the checkbox) to an integer (1 if True, 0 if False)
     luxurious_value = 1 if luxurious else 0
 
-    # Create the input vector (ensure the order of features matches your model training)
+    # Create the input vector with the new "luxurious" attribute as the last feature
     input_features = np.array([
         rooms,
         area,
@@ -146,11 +154,14 @@ def predict_apartment(rooms, area, town, luxurious):
         df['tax_income'].iloc[0],
         luxurious_value
     ])
-    
-    input_features = input_features.reshape(1, 8)  # Now 8 features are expected
-    prediction = model.predict(input_features)
-    return np.round(prediction[0][0], 0)
+    input_features = input_features.reshape(1, 8)
+    prediction = random_forest_model.predict(input_features)
+    return np.round(prediction[0], 0)
+
+# %%
+print(predict_apartment(3, 100, 'Zürich', True))
 
+# %%
 # Create the Gradio interface with an extra input for luxurious (yes/no)
 iface = gr.Interface(
     fn=predict_apartment,
@@ -160,7 +171,13 @@ iface = gr.Interface(
         gr.Dropdown(choices=list(locations.keys()), label="Town", type="value"),
         gr.Checkbox(label="Luxurious?")
     ],
-    outputs=gr.Number()
+    outputs=gr.Number(),
+    examples=[
+        [4.5, 120, "Dietikon", True],
+        [3.5, 60, "Winterthur", False]
+    ]
 )
 
-iface.launch()
+iface.launch()
+
+