import gradio as gr
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.svm import SVC
from sklearn.preprocessing import StandardScaler

# --- Sample Morningstar-style data ---
data = {
    "5Y_Return": [14.0, 7.5, 13.2, 6.0, 15.0, 8.0, 12.0, 6.5, 10.5, 7.2],
    "Volatility": [8.0, 6.5, 7.8, 9.0, 7.0, 6.2, 7.1, 8.5, 6.8, 7.9],
    "Risk_Score": [2, 3, 2, 4, 1, 3, 2, 4, 2, 3],
    "Rating": ["Good", "Bad", "Good", "Bad", "Good", "Bad", "Good", "Bad", "Good", "Bad"]
}
df = pd.DataFrame(data)
df["Label"] = df["Rating"].map({"Good": 1, "Bad": 0})

# --- Train full SVM model for prediction ---
X = df[["5Y_Return", "Volatility", "Risk_Score"]]
y = df["Label"]

scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)

model = SVC(kernel="linear", probability=True)
model.fit(X_scaled, y)

# --- Function to classify and plot 2D SVM boundary ---
def classify_and_plot(return_5y, volatility, risk_score):
    # Predict
    input_data = [[return_5y, volatility, risk_score]]
    input_scaled = scaler.transform(input_data)
    prediction = model.predict(input_scaled)[0]
    confidence = model.predict_proba(input_scaled)[0][prediction]
    result = "Good Investment" if prediction == 1 else "Bad Investment"

    # For plotting, use only 2D
    X_2d = df[["5Y_Return", "Volatility"]].values
    y_2d = df["Label"].values
    scaler_2d = StandardScaler()
    X_2d_scaled = scaler_2d.fit_transform(X_2d)

    model_2d = SVC(kernel="linear")
    model_2d.fit(X_2d_scaled, y_2d)

    # Plot decision boundary
    fig, ax = plt.subplots(figsize=(6, 5))
    ax.scatter(X_2d_scaled[:, 0], X_2d_scaled[:, 1], c=y_2d, cmap="bwr", edgecolors="k", s=60)

    # Support vectors
    ax.scatter(model_2d.support_vectors_[:, 0], model_2d.support_vectors_[:, 1],
               s=150, facecolors='none', edgecolors='k', linewidths=1.5, label="Support Vectors")

    # Decision boundary
    xlim = ax.get_xlim()
    ylim = ax.get_ylim()
    xx = np.linspace(xlim[0], xlim[1], 30)
    yy = np.linspace(ylim[0], ylim[1], 30)
    YY, XX = np.meshgrid(yy, xx)
    xy = np.vstack([XX.ravel(), YY.ravel()]).T
    Z = model_2d.decision_function(xy).reshape(XX.shape)
    ax.contour(XX, YY, Z, colors='k', levels=[-1, 0, 1],
               alpha=0.7, linestyles=['--', '-', '--'])

    # Annotations
    ax.set_title("SVM Decision Boundary (2 Features)")
    ax.set_xlabel("5Y Return (scaled)")
    ax.set_ylabel("Volatility (scaled)")
    ax.legend()
    ax.grid(True)

    # Save and return image
    plot_path = "/tmp/svm_plot.png"
    fig.savefig(plot_path)
    plt.close(fig)

    return f"{result} (Confidence: {confidence:.2f})", plot_path

# --- Gradio UI ---
with gr.Blocks() as demo:
    gr.Markdown("## 🧠 SVM Classifier: Mutual Fund Recommendation")

    with gr.Row():
        return_input = gr.Number(label="5-Year Return (%)", value=10.0)
        vol_input = gr.Number(label="Volatility (%)", value=7.0)
        risk_input = gr.Number(label="Risk Score (1=Low, 5=High)", value=3)

    classify_btn = gr.Button("Classify and Show Decision Boundary")
    output_label = gr.Textbox(label="Prediction")

    gr.Markdown("""### 📊 Benchmark Guide  
**🔴 Blue Dots = Good Investments**  
**🔴 Red Dots = Bad Investments**  
**⚫ Solid Black Line = Decision Boundary**  
**⚫ Dashed Lines = Margins (distance to support vectors)**  
**⭕ Large Hollow Dots = Support Vectors (key data points)**  
""")

    output_plot = gr.Image(label="SVM Decision Boundary")

    classify_btn.click(
        fn=classify_and_plot,
        inputs=[return_input, vol_input, risk_input],
        outputs=[output_label, output_plot]
    )

if __name__ == "__main__":
    demo.launch()