Update app.py

app.py

@@ -5,23 +5,15 @@ import torch
 import json
 import pickle
 from pytorch_tabular import TabularModel
-from sklearn.preprocessing import LabelEncoder
+
 
 # Load model
 model = TabularModel.load_model("FTTransformerModel")
 
-# Wrap datamodule.label_encoder into a list if it's a single LabelEncoder
-dm = model.datamodule
-if hasattr(dm, "label_encoder") and isinstance(dm.label_encoder, LabelEncoder):
-    dm.label_encoder = [dm.label_encoder]
-
 # Load threshold
 with open("FTTransformerModel/threshold.json", "r") as f:
     threshold = json.load(f)["threshold"]
 
-# Load encoders
-with open("FTTransformerModel/encoder.pkl", "rb") as f:
-    label_encoders = pickle.load(f)
 
 # User input
 st.title("Cardiovascular Disease Risk Prediction")
@@ -32,8 +24,8 @@ height = st.number_input("Height (cm)", min_value=100, max_value=250, value=170)
 weight = st.number_input("Weight (kg)", min_value=30, max_value=200, value=70)
 systolic = st.number_input("Systolic BP", min_value=80, max_value=250, value=120)
 diastolic = st.number_input("Diastolic BP", min_value=40, max_value=150, value=80)
-cholesterol = st.selectbox("Cholesterol", [1, 2, 3], format_func=lambda x: {1: "Normal", 2: "High", 3: "Very High"}[x])
-gluc = st.selectbox("Glucose", [1, 2, 3], format_func=lambda x: {1: "Normal", 2: "High", 3: "Very High"}[x])
+cholesterol = st.selectbox("Cholesterol", [0, 1, 2], format_func=lambda x: {0: "Normal", 1: "High", 2: "Very High"}[x])
+gluc = st.selectbox("Glucose", [0, 1, 2], format_func=lambda x: {0: "Normal", 1: "High", 2: "Very High"}[x])
 gender = st.selectbox("Gender", [0, 1], format_func=lambda x: {0: "Female", 1: "Male"}[x])
 smoke = st.selectbox("Do you smoke?", [0, 1], format_func=lambda x: "Yes" if x else "No")
 alco = st.selectbox("Do you drink alcohol?", [0, 1], format_func=lambda x: "Yes" if x else "No")
@@ -53,10 +45,14 @@ input_data = pd.DataFrame([{
     "active": active
 }])
 
-categorical_cols = ['gender', 'cholesterol', 'gluc', 'smoke', 'alco', 'active']
-for col in categorical_cols:
-    le = label_encoders[col]
-    input_data[col] = le.transform(input_data[col])
+
+input_data["bmi"] = input_data["weight"] / ((input_data["height"]/100)**2)
+
+input_data["pulse_pressure"] = input_data["ap_hi"] - input_data["ap_lo"]
+
+input_data["hypertension"] = (
+    (input_data["ap_hi"] > 140) | (input_data["ap_lo"] > 90)
+).astype(int)
 
 if st.button("Predict CVD Risk"):
     preds = model.predict(input_data)