Update app.py

app.py

@@ -1,147 +1,124 @@
-import json
-import joblib
-import numpy as np
-import pandas as pd
+import json, joblib, pandas as pd, numpy as np
 import gradio as gr
+import seaborn as sns
+import matplotlib.pyplot as plt
 
-# ---------- Load artifacts ----------
-MODEL_PATH = "maternal_rf_model.joblib"
-META_PATH = "maternal_metadata.json"
-DATA_PATH = "maternal_cleaned.csv"  # optional, for example defaults or sanity checks
-
-model = joblib.load(MODEL_PATH)
-
-with open(META_PATH, "r", encoding="utf-8") as f:
+# Load model + metadata + dataset
+model = joblib.load("maternal_rf_model.joblib")
+with open("maternal_metadata.json","r",encoding="utf-8") as f:
     meta = json.load(f)
+df_clean = pd.read_csv("maternal_cleaned.csv")
 
 numeric_features = meta["numeric_features"]
 categorical_features = meta["categorical_features"]
-target_col = meta["target"]
-
-# Optional: load cleaned dataset to compute sensible defaults/ranges
-try:
-    df_clean = pd.read_csv(DATA_PATH)
-except Exception:
-    df_clean = None
-
-# ---------- Define categorical options ----------
-# Ensure these match your training preprocessing categories
-ANAEMIA_OPTS = ["None", "Minimal", "Medium", "Higher"]
-JAUNDICE_OPTS = ["None", "Minimal", "Medium"]
-FETAL_POSITION_OPTS = ["Normal", "Abnormal"]
-FETAL_MOVEMENT_OPTS = ["Yes", "No"]
-URINE_ALBUMIN_OPTS = ["Negative", "Positive"]
-URINE_SUGAR_OPTS = ["Negative", "Positive"]
-
-# ---------- Defaults from dataset (median or most frequent) ----------
-def default_num(name, fallback=0.0):
-    if df_clean is not None and name in df_clean.columns:
-        return float(np.nanmedian(df_clean[name].values))
-    return float(fallback)
-
-def default_cat(name, options, fallback=None):
-    if df_clean is not None and name in df_clean.columns:
-        mode = df_clean[name].dropna().astype(str).mode()
-        if len(mode) > 0 and mode[0] in options:
-            return mode[0]
-    return fallback or options[0]
-
-DEFAULTS = {
-    "Age": default_num("Age", 22),
-    "Gravida": default_num("Gravida", 1),
-    "GestationWeeks": default_num("GestationWeeks", 30),
-    "WeightKg": default_num("WeightKg", 56),
-    "HeightCm": default_num("HeightCm", 160),
-    "BP_Systolic": default_num("BP_Systolic", 100),
-    "BP_Diastolic": default_num("BP_Diastolic", 60),
-    "FetalHR": default_num("FetalHR", 140),
-    "Anaemia": default_cat("Anaemia", ANAEMIA_OPTS, "None"),
-    "Jaundice": default_cat("Jaundice", JAUNDICE_OPTS, "None"),
-    "FetalPosition": default_cat("FetalPosition", FETAL_POSITION_OPTS, "Normal"),
-    "FetalMovement": default_cat("FetalMovement", FETAL_MOVEMENT_OPTS, "Yes"),
-    "UrineAlbumin": default_cat("UrineAlbumin", URINE_ALBUMIN_OPTS, "Negative"),
-    "UrineSugar": default_cat("UrineSugar", URINE_SUGAR_OPTS, "Negative"),
-}
 
 # ---------- Prediction function ----------
-def predict_risk(
-    age, gravida, gest_weeks, weight, height_cm,
-    bp_sys, bp_dias, fetal_hr,
-    anaemia, jaundice, fetal_position, fetal_movement, urine_albumin, urine_sugar
-):
-    # Build a single-row DataFrame with exact column order
+def predict_risk(age, gravida, gest_weeks, weight, height_cm,
+                 bp_sys, bp_dias, fetal_hr,
+                 anaemia, jaundice, fetal_position, fetal_movement,
+                 urine_albumin, urine_sugar):
     row = {
-        "Age": age,
-        "Gravida": gravida,
-        "GestationWeeks": gest_weeks,
-        "WeightKg": weight,
-        "HeightCm": height_cm,
-        "BP_Systolic": bp_sys,
-        "BP_Diastolic": bp_dias,
+        "Age": age, "Gravida": gravida, "GestationWeeks": gest_weeks,
+        "WeightKg": weight, "HeightCm": height_cm,
+        "BP_Systolic": bp_sys, "BP_Diastolic": bp_dias,
         "FetalHR": fetal_hr,
-        "Anaemia": anaemia,
-        "Jaundice": jaundice,
-        "FetalPosition": fetal_position,
-        "FetalMovement": fetal_movement,
-        "UrineAlbumin": urine_albumin,
-        "UrineSugar": urine_sugar,
+        "Anaemia": anaemia, "Jaundice": jaundice,
+        "FetalPosition": fetal_position, "FetalMovement": fetal_movement,
+        "UrineAlbumin": urine_albumin, "UrineSugar": urine_sugar
     }
-    X = pd.DataFrame([row], columns=numeric_features + categorical_features)
-
-    # Predict
-    prob = None
-    try:
-        prob = model.predict_proba(X)[:, 1][0]
-    except Exception:
-        # If model lacks predict_proba (shouldn’t happen for RandomForest), fallback
-        prob = float(model.predict(X)[0])
-
+    X = pd.DataFrame([row], columns=numeric_features+categorical_features)
+    prob = model.predict_proba(X)[:,1][0]
     pred = int(model.predict(X)[0])
-    label = "High Risk" if pred == 1 else "Not High Risk"
-
-    # Friendly output with rounded probability
-    return {
-        "Prediction": label,
-        "Probability_high_risk": round(float(prob), 4)
-    }
+    label = "High Risk" if pred==1 else "Not High Risk"
+    return {"Prediction":label,"Probability_high_risk":round(prob,4)}
+
+# ---------- Plot functions ----------
+def plot_age_distribution():
+    fig, ax = plt.subplots(figsize=(6,4))
+    sns.histplot(df_clean["Age"], bins=10, kde=True, ax=ax, color="skyblue")
+    ax.set_title("Age Distribution")
+    return fig
+
+def plot_risk_counts():
+    fig, ax = plt.subplots(figsize=(6,4))
+    sns.countplot(x="HighRisk", data=df_clean, ax=ax, palette="Set2")
+    ax.set_title("High Risk vs Non-Risk Counts")
+    return fig
+
+def plot_gestation_box():
+    fig, ax = plt.subplots(figsize=(6,4))
+    sns.boxplot(x="HighRisk", y="GestationWeeks", data=df_clean, ax=ax, palette="Set2")
+    ax.set_title("Gestation Weeks vs Risk")
+    return fig
+
+def plot_feature_importance():
+    ohe = model.named_steps["preprocessor"].named_transformers_["cat"].named_steps["onehot"]
+    cat_names = ohe.get_feature_names_out(categorical_features)
+    feature_names = numeric_features + list(cat_names)
+    importances = model.named_steps["clf"].feature_importances_
+    feat_imp = pd.DataFrame({"Feature":feature_names,"Importance":importances})
+    feat_imp = feat_imp.sort_values("Importance",ascending=False).head(10)
+    fig, ax = plt.subplots(figsize=(8,5))
+    sns.barplot(x="Importance", y="Feature", data=feat_imp, ax=ax, palette="viridis")
+    ax.set_title("Top 10 Feature Importances")
+    return fig
+
+def plot_corr_heatmap():
+    fig, ax = plt.subplots(figsize=(8,6))
+    corr = df_clean[numeric_features+["HighRisk"]].corr()
+    sns.heatmap(corr, annot=True, cmap="coolwarm", fmt=".2f", ax=ax)
+    ax.set_title("Correlation Heatmap")
+    return fig
 
 # ---------- Gradio UI ----------
 with gr.Blocks(title="Maternal Risk Prediction") as demo:
-    gr.Markdown(
-        "## Maternal Risk Prediction\n"
-        "Enter clinical inputs to estimate high-risk pregnancy likelihood. "
-        "This tool uses a trained RandomForest model."
-    )
-
-    with gr.Row():
-        with gr.Column():
-            age_in = gr.Number(label="Age (years)", value=DEFAULTS["Age"])
-            gravida_in = gr.Number(label="Gravida (1/2/3)", value=DEFAULTS["Gravida"])
-            gest_in = gr.Number(label="Gestation Weeks", value=DEFAULTS["GestationWeeks"])
-            weight_in = gr.Number(label="Weight (kg)", value=DEFAULTS["WeightKg"])
-            height_in = gr.Number(label="Height (cm)", value=DEFAULTS["HeightCm"])
-        with gr.Column():
-            bp_sys_in = gr.Number(label="BP Systolic (mmHg)", value=DEFAULTS["BP_Systolic"])
-            bp_dias_in = gr.Number(label="BP Diastolic (mmHg)", value=DEFAULTS["BP_Diastolic"])
-            fetal_hr_in = gr.Number(label="Fetal Heart Rate (bpm)", value=DEFAULTS["FetalHR"])
-            anaemia_in = gr.Dropdown(ANAEMIA_OPTS, label="Anaemia", value=DEFAULTS["Anaemia"])
-            jaundice_in = gr.Dropdown(JAUNDICE_OPTS, label="Jaundice", value=DEFAULTS["Jaundice"])
-        with gr.Column():
-            fetal_pos_in = gr.Dropdown(FETAL_POSITION_OPTS, label="Fetal Position", value=DEFAULTS["FetalPosition"])
-            fetal_mov_in = gr.Dropdown(FETAL_MOVEMENT_OPTS, label="Fetal Movement", value=DEFAULTS["FetalMovement"])
-            urine_alb_in = gr.Dropdown(URINE_ALBUMIN_OPTS, label="Urine Albumin", value=DEFAULTS["UrineAlbumin"])
-            urine_sug_in = gr.Dropdown(URINE_SUGAR_OPTS, label="Urine Sugar", value=DEFAULTS["UrineSugar"])
-
-    predict_btn = gr.Button("Predict Risk")
-
-    out_json = gr.JSON(label="Result")
-
-    predict_btn.click(
-        predict_risk,
-        inputs=[age_in, gravida_in, gest_in, weight_in, height_in,
-                bp_sys_in, bp_dias_in, fetal_hr_in,
-                anaemia_in, jaundice_in, fetal_pos_in, fetal_mov_in, urine_alb_in, urine_sug_in],
-        outputs=[out_json]
-    )
+    gr.Markdown("## Maternal Risk Prediction Dashboard")
+
+    with gr.Tab("Prediction"):
+        gr.Markdown("Enter maternal health parameters to predict risk.")
+        with gr.Row():
+            age = gr.Number(label="Age")
+            gravida = gr.Number(label="Gravida")
+            gest = gr.Number(label="Gestation Weeks")
+            weight = gr.Number(label="Weight (kg)")
+            height = gr.Number(label="Height (cm)")
+        with gr.Row():
+            bp_sys = gr.Number(label="BP Systolic")
+            bp_dias = gr.Number(label="BP Diastolic")
+            fetal_hr = gr.Number(label="Fetal Heart Rate")
+            anaemia = gr.Dropdown(["None","Minimal","Medium","Higher"], label="Anaemia")
+            jaundice = gr.Dropdown(["None","Minimal","Medium"], label="Jaundice")
+        with gr.Row():
+            fetal_pos = gr.Dropdown(["Normal","Abnormal"], label="Fetal Position")
+            fetal_mov = gr.Dropdown(["Yes","No"], label="Fetal Movement")
+            urine_alb = gr.Dropdown(["Negative","Positive"], label="Urine Albumin")
+            urine_sug = gr.Dropdown(["Negative","Positive"], label="Urine Sugar")
+        out = gr.JSON(label="Result")
+        btn = gr.Button("Predict Risk")
+        btn.click(predict_risk,
+                  inputs=[age,gravida,gest,weight,height,
+                          bp_sys,bp_dias,fetal_hr,
+                          anaemia,jaundice,fetal_pos,fetal_mov,urine_alb,urine_sug],
+                  outputs=out)
+
+    with gr.Tab("Data Insights"):
+        gr.Markdown("### Dataset Overview")
+        gr.Plot(plot_age_distribution)
+        gr.Plot(plot_risk_counts)
+        gr.Plot(plot_gestation_box)
+
+    with gr.Tab("Model Insights"):
+        gr.Markdown("### Model Behavior")
+        gr.Plot(plot_feature_importance)
+        gr.Plot(plot_corr_heatmap)
+
+    with gr.Tab("About"):
+        gr.Markdown("""
+        ### About this App
+        This dashboard predicts maternal high-risk pregnancy using a RandomForest model.
+        - **Dataset:** Cleaned maternal health records
+        - **Features:** Age, Gravida, Gestation Weeks, Weight, Height, BP, Fetal HR, Anaemia, Jaundice, Fetal Position, Fetal Movement, Urine Albumin, Urine Sugar
+        - **Output:** High Risk vs Not High Risk with probability
+        """)
 
 demo.launch()