app.py

import gradio as gr
import pandas as pd
import tempfile

# --------------------------------------------------
# Metric Computation Function
# --------------------------------------------------
def compute_metrics(tp, tn, fp, fn):

    tp, tn, fp, fn = int(tp), int(tn), int(fp), int(fn)

    total = tp + tn + fp + fn

    def safe_div(a, b):
        return a / b if b != 0 else 0

    accuracy = safe_div(tp + tn, total)
    precision = safe_div(tp, tp + fp)
    recall = safe_div(tp, tp + fn)              # Sensitivity
    specificity = safe_div(tn, tn + fp)
    f1_score = safe_div(2 * precision * recall, precision + recall)
    npv = safe_div(tn, tn + fn)
    fpr = safe_div(fp, fp + tn)
    fnr = safe_div(fn, fn + tp)
    fdr = safe_div(fp, fp + tp)
    balanced_accuracy = (recall + specificity) / 2

    metrics = {
        "Accuracy": accuracy,
        "Precision (PPV)": precision,
        "Recall / Sensitivity (TPR)": recall,
        "Specificity (TNR)": specificity,
        "F1 Score": f1_score,
        "Negative Predictive Value (NPV)": npv,
        "False Positive Rate (FPR)": fpr,
        "False Negative Rate (FNR)": fnr,
        "False Discovery Rate (FDR)": fdr,
        "Balanced Accuracy": balanced_accuracy,
        "Total Samples": total
    }

    df = pd.DataFrame(list(metrics.items()), columns=["Metric", "Value"])

    # Save CSV temporarily for download
    temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".csv")
    df.to_csv(temp_file.name, index=False)

    return df, temp_file.name


# --------------------------------------------------
# Gradio Interface
# --------------------------------------------------
with gr.Blocks(title="Classification Metrics Calculator") as demo:

    gr.Markdown(
        """
        # 📊 Classification Metrics Calculator
        Enter Confusion Matrix values (TP, TN, FP, FN) to compute evaluation metrics.
        """
    )

    with gr.Row():
        tp = gr.Number(value=50, label="True Positives (TP)")
        tn = gr.Number(value=40, label="True Negatives (TN)")

    with gr.Row():
        fp = gr.Number(value=10, label="False Positives (FP)")
        fn = gr.Number(value=5, label="False Negatives (FN)")

    compute_btn = gr.Button("Compute Metrics")

    output_table = gr.Dataframe(
        headers=["Metric", "Value"],
        datatype=["str", "number"],
        label="Computed Metrics"
    )

    download_file = gr.File(label="⬇ Download Results (CSV)")

    compute_btn.click(
        fn=compute_metrics,
        inputs=[tp, tn, fp, fn],
        outputs=[output_table, download_file]
    )

# Launch (important for HF Spaces)
demo.launch()