water_quality_index.py

import streamlit as st
import pandas as pd
import numpy as np
import joblib
import plotly.express as px
import base64
from sklearn.preprocessing import LabelEncoder

# === Thresholds for Rule-Based Classification ===
thresholds = {
    'pH_min': 6.0, 'pH_max': 9.0,
    'BOD': 3.0,
    'COD': 25.0,
    'TSS': 50.0,
    'DO': 4.0,
    'Nitrate': 10.0,
    'Phosphate': 0.2,
    'FecalColiform': 1000
}

def categorize_sample(row):
    pH = row['pH (Potential Hydrogen)']
    BOD = row['BOD (Biological Oxygen Demand) (mg/L)']
    COD = row['COD (Chemical Oxygen Demand) (mg/L)']
    DO = row['DO (Dissolved Oxygen) (mg/L)']
    nitrate = row['NO3N (Nitrat) (mg/L)']
    phosphate = row['Total Phosphat (mg/L)']
    fecal = row['Fecal Coliform (MPN/100 mL)']
    TSS = row['TSS (Total Suspended Solid) (mg/L)']

    if (
        thresholds['pH_min'] <= pH <= thresholds['pH_max'] and
        BOD <= thresholds['BOD'] and
        COD <= thresholds['COD'] and
        DO >= thresholds['DO'] and
        nitrate <= thresholds['Nitrate'] and
        phosphate <= thresholds['Phosphate'] and
        fecal <= thresholds['FecalColiform'] and
        TSS <= thresholds['TSS']
    ):
        return "Safe", "Safe"

    categories = []
    if COD > thresholds['COD'] * 1.5 or pH < thresholds['pH_min'] or pH > thresholds['pH_max'] or TSS > thresholds['TSS']:
        categories.append("Chemical")
    if BOD > thresholds['BOD'] or DO < thresholds['DO'] or fecal > thresholds['FecalColiform'] or TSS > thresholds['TSS']:
        categories.append("Biological")
    if nitrate > thresholds['Nitrate'] or phosphate > thresholds['Phosphate'] or TSS > thresholds['TSS']:
        categories.append("Eutrophication")

    priority_order = ["Chemical", "Biological", "Eutrophication"]
    for cat in priority_order:
        if cat in categories:
            return ", ".join(categories), cat

    return "Safe", "Safe"

# === Streamlit App ===
def run():
    svc_model = joblib.load("svc_model.pkl")
    xgb_model = joblib.load("xgb_model.pkl")
    imputer = joblib.load("imputer.pkl")
    scaler = joblib.load("scaler.pkl")
    label_encoder = joblib.load("label_encoder.pkl")

    feature_cols = [
        "pH (Potential Hydrogen)",
        "BOD (Biological Oxygen Demand) (mg/L)",
        "COD (Chemical Oxygen Demand) (mg/L)",
        "TSS (Total Suspended Solid) (mg/L)",
        "DO (Dissolved Oxygen) (mg/L)",
        "NO3N (Nitrat) (mg/L)",
        "Total Phosphat (mg/L)",
        "Fecal Coliform (MPN/100 mL)"
    ]

    st.set_page_config(page_title="Water Quality Classifier Dashboard", layout="wide")
    st.title("💧 Water Quality Prediction and Rule-Based Evaluation")

    model_choice = st.selectbox("Select Model", ["SVC + SMOTETomek", "XGBoost + SMOTETomek"])
    model = svc_model if model_choice == "SVC + SMOTETomek" else xgb_model

    st.header("📥 Input Data")
    data_option = st.radio("Choose Input Method", ["Upload CSV", "Manual Entry"])
    input_df = None

    if data_option == "Upload CSV":
        uploaded_file = st.file_uploader("Upload your CSV file", type=["csv"])
        if uploaded_file:
            df = pd.read_csv(uploaded_file)
            missing_cols = [col for col in feature_cols if col not in df.columns]
            if missing_cols:
                st.error(f"Missing required columns: {missing_cols}")
            else:
                input_df = df[feature_cols]
    else:
        with st.form("manual_form"):
            ph = st.number_input("pH", min_value=0.0, max_value=14.0, value=7.0)
            bod = st.number_input("BOD (mg/L)", min_value=0.0, max_value=100.0, value=2.0)
            cod = st.number_input("COD (mg/L)", min_value=0.0, max_value=500.0, value=10.0)
            tss = st.number_input("TSS (mg/L)", min_value=0.0, max_value=1000.0, value=20.0)
            do = st.number_input("DO (mg/L)", min_value=0.0, max_value=20.0, value=5.0)
            no3 = st.number_input("NO3N (mg/L)", min_value=0.0, max_value=10.0, value=1.0)
            tp = st.number_input("Total Phosphat (mg/L)", min_value=0.0, max_value=10.0, value=0.1)
            fecal = st.number_input("Fecal Coliform (MPN/100 mL)", min_value=0.0, max_value=1000000.0, value=500.0)
            submitted = st.form_submit_button("Predict")

        if submitted:
            input_df = pd.DataFrame([{
                "pH (Potential Hydrogen)": ph,
                "BOD (Biological Oxygen Demand) (mg/L)": bod,
                "COD (Chemical Oxygen Demand) (mg/L)": cod,
                "TSS (Total Suspended Solid) (mg/L)": tss,
                "DO (Dissolved Oxygen) (mg/L)": do,
                "NO3N (Nitrat) (mg/L)": no3,
                "Total Phosphat (mg/L)": tp,
                "Fecal Coliform (MPN/100 mL)": fecal
            }])

    if input_df is not None:
        st.header("🔍 Prediction Results")

        try:
            X_imp = imputer.transform(input_df)
            X_scaled = scaler.transform(X_imp)
            y_proba = model.predict_proba(X_scaled)
            y_pred = model.predict(X_scaled)
            pred_class = label_encoder.inverse_transform(y_pred)[0]

            # Rule-Based Evaluation
            rule_violations, rule_label = categorize_sample(input_df.iloc[0])

            # Display results
            st.markdown(f"### 🧪 ML Predicted Class: `{pred_class}`")
            st.markdown(f"### 📏 Rule-Based Class: `{rule_label}`")
            st.markdown(f"**Violations Detected:** {rule_violations}")

            fig_pie = px.pie(
                names=label_encoder.classes_,
                values=y_proba[0],
                title="Prediction Probability per Class",
                color_discrete_sequence=px.colors.qualitative.Set3
            )
            st.plotly_chart(fig_pie, use_container_width=True)

            # Export Results
            input_df["Predicted Class (ML)"] = pred_class
            input_df["Rule-Based Class"] = rule_label
            input_df["Rule-Based Violations"] = rule_violations
            input_df[[f"Prob_{cls}" for cls in label_encoder.classes_]] = y_proba
            csv = input_df.to_csv(index=False)
            b64 = base64.b64encode(csv.encode()).decode()
            href = f'<a href="data:file/csv;base64,{b64}" download="prediction_result.csv">Download CSV File</a>'
            st.subheader("📤 Download Result")
            st.markdown(href, unsafe_allow_html=True)

        except Exception as e:
            st.error(f"Prediction failed: {e}")

    st.markdown("---")
    st.caption("Developed with ❤️ for integrated ML + expert rule water quality system")