ranimeree commited on
Commit
2db51b2
·
verified ·
1 Parent(s): 91f5848

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +50 -58
app.py CHANGED
@@ -2,74 +2,66 @@ import gradio as gr
2
  import pandas as pd
3
  import numpy as np
4
  import pickle
5
- from sklearn.compose import ColumnTransformer
6
- from sklearn.preprocessing import StandardScaler, OneHotEncoder
7
- from sklearn.pipeline import Pipeline
8
- from sklearn.impute import SimpleImputer
9
 
10
- def transformer_fn():
11
- """
12
- Returns a preprocessing pipeline that matches the training approach.
13
- """
14
- # Define numeric and categorical features
15
  numeric_features = ['age', 'avg_glucose_level', 'bmi']
16
- categorical_features = [
17
- 'gender', 'hypertension', 'heart_disease', 'ever_married',
18
- 'work_type', 'Residence_type', 'smoking_status'
19
- ]
20
 
21
- # Create numeric pipeline
22
- numeric_transformer = Pipeline(steps=[
23
- ('imputer', SimpleImputer(strategy='median')),
24
- ('scaler', StandardScaler())
25
- ])
26
 
27
- # Create categorical pipeline
28
- categorical_transformer = Pipeline(steps=[
29
- ('imputer', SimpleImputer(strategy='constant', fill_value='missing')),
30
- ('onehot', OneHotEncoder(drop='first', sparse=False))
31
- ])
32
 
33
- # Combine preprocessing steps
34
- preprocessor = ColumnTransformer(
35
- transformers=[
36
- ('num', numeric_transformer, numeric_features),
37
- ('cat', categorical_transformer, categorical_features)
38
- ],
39
- remainder='drop' # Drop any other columns
40
- )
 
 
41
 
42
- return preprocessor
43
-
44
- # Load the model
45
- with open('model.pkl', 'rb') as f:
46
- model = pickle.load(f)
47
-
48
- # Get the transformer
49
- transformer = transformer_fn()
50
 
51
- def predict(gender, age, hypertension, ever_married, work_type, heart_disease, avg_glucose_level, bmi, smoking_status, Residence_type):
52
- # Create input data
 
 
53
  input_data = {
54
- 'gender': [gender],
55
- 'age': [float(age)], # Convert to float as expected by the model
56
- 'hypertension': [int(hypertension == 'Yes')], # Convert to int
57
- 'heart_disease': [int(heart_disease == 'Yes')],
58
- 'ever_married': [ever_married],
59
- 'work_type': [work_type],
60
- 'Residence_type': [Residence_type],
61
- 'avg_glucose_level': [float(avg_glucose_level)],
62
- 'bmi': [float(bmi)],
63
- 'smoking_status': [smoking_status]
64
  }
65
-
66
- # Convert to DataFrame
67
- input_df = pd.DataFrame(input_data)
68
-
69
- # Transform the input using the same transformer used during training
70
  try:
71
- X_transformed = transformer.fit_transform(input_df) # Using fit_transform since it's a new transformer instance
72
- prediction = model.predict_proba(X_transformed)[0][1]
 
73
  return f"The probability of stroke is {prediction:.2%}"
74
  except Exception as e:
75
  return f"Error making prediction: {str(e)}"
 
2
  import pandas as pd
3
  import numpy as np
4
  import pickle
5
+ from sklearn.preprocessing import StandardScaler
 
 
 
6
 
7
+ def preprocess_input(data_dict):
8
+ """Preprocess input data to match the training format"""
9
+ df = pd.DataFrame([data_dict])
10
+
11
+ # Numeric features
12
  numeric_features = ['age', 'avg_glucose_level', 'bmi']
 
 
 
 
13
 
14
+ # Scale numeric features
15
+ scaler = StandardScaler()
16
+ df[numeric_features] = scaler.fit_transform(df[numeric_features])
 
 
17
 
18
+ # Create dummy variables for categorical features
19
+ df = pd.get_dummies(df, columns=['gender', 'hypertension', 'heart_disease',
20
+ 'ever_married', 'work_type', 'Residence_type',
21
+ 'smoking_status'])
 
22
 
23
+ # Ensure all expected columns are present
24
+ expected_columns = [
25
+ 'num__age', 'num__avg_glucose_level', 'num__bmi',
26
+ 'cat__gender_Male', 'cat__gender_Other', 'cat__hypertension_1',
27
+ 'cat__heart_disease_1', 'cat__ever_married_Yes',
28
+ 'cat__work_type_Never_worked', 'cat__work_type_Private',
29
+ 'cat__work_type_Self-employed', 'cat__work_type_children',
30
+ 'cat__Residence_type_Urban', 'cat__smoking_status_formerly smoked',
31
+ 'cat__smoking_status_never smoked', 'cat__smoking_status_smokes'
32
+ ]
33
 
34
+ for col in expected_columns:
35
+ if col not in df.columns:
36
+ df[col] = 0
37
+
38
+ return df[expected_columns]
 
 
 
39
 
40
+ def predict(gender, age, hypertension, ever_married, work_type, heart_disease,
41
+ avg_glucose_level, bmi, smoking_status, Residence_type):
42
+ """Make prediction using the loaded model"""
43
+ # Create input dictionary
44
  input_data = {
45
+ 'gender': gender,
46
+ 'age': age,
47
+ 'hypertension': 1 if hypertension == 'Yes' else 0,
48
+ 'heart_disease': 1 if heart_disease == 'Yes' else 0,
49
+ 'ever_married': ever_married,
50
+ 'work_type': work_type,
51
+ 'Residence_type': Residence_type,
52
+ 'avg_glucose_level': avg_glucose_level,
53
+ 'bmi': bmi,
54
+ 'smoking_status': smoking_status
55
  }
56
+
57
+ # Preprocess the input
58
+ processed_input = preprocess_input(input_data)
59
+
60
+ # Load and use the model
61
  try:
62
+ with open('model.pkl', 'rb') as file:
63
+ model = pickle.load(file)
64
+ prediction = model.predict_proba(processed_input)[0][1]
65
  return f"The probability of stroke is {prediction:.2%}"
66
  except Exception as e:
67
  return f"Error making prediction: {str(e)}"