app.py

# app.py
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from scipy import stats
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, roc_auc_score
from tabulate import tabulate
import warnings
import traceback
import gradio as gr
import os
import git

# --- Main Class (Slightly Refactored for Interactivity) ---
# The core logic remains, but we separate data loading from analysis functions.
warnings.filterwarnings('ignore')
plt.style.use('default')
sns.set_palette("husl")


class EnhancedAIvsRealGazeAnalyzer:
    def __init__(self):
        self.questions = ['Q1', 'Q2', 'Q3', 'Q4', 'Q5', 'Q6']
        self.correct_answers = {'Pair1': 'B', 'Pair2': 'B', 'Pair3': 'B', 'Pair4': 'B', 'Pair5': 'B', 'Pair6': 'B'}
        self.combined_data = None
        self.response_data = None
        self.numeric_cols = []
        self.time_metrics = []

    def load_and_process_data(self, base_path, response_file):
        """Loads all data and preprocesses it once."""
        print("Loading and processing data...")
        # Load response data
        self.response_data = pd.read_excel(response_file) if response_file.endswith('.xlsx') else pd.read_csv(
            response_file)
        self.response_data.columns = self.response_data.columns.str.strip()
        for pair, correct_answer in self.correct_answers.items():
            if pair in self.response_data.columns:
                self.response_data[f'{pair}_Correct'] = (
                            self.response_data[pair].astype(str).str.strip().str.upper() == correct_answer)

        # Load eye-tracking data
        all_data = {}
        for question in self.questions:
            file_path = f"{base_path}/Filtered_GenAI_Metrics_cleaned_{question}.xlsx"
            if os.path.exists(file_path):
                xls = pd.ExcelFile(file_path)
                all_data[question] = {sheet_name: pd.read_excel(xls, sheet_name) for sheet_name in xls.sheet_names}

        # Combine and merge
        all_dfs = [df.copy().assign(Question=q, Metric_Type=m) for q, qd in all_data.items() for m, df in qd.items()]
        if not all_dfs:
            raise ValueError("No eye-tracking data files were found or loaded.")

        self.combined_data = pd.concat(all_dfs, ignore_index=True)
        self.combined_data.columns = self.combined_data.columns.str.strip()

        # Merge with responses
        et_id_col = next((c for c in self.combined_data.columns if 'participant' in c.lower()), None)
        resp_id_col = next((c for c in self.response_data.columns if 'participant' in c.lower()), None)

        response_long = self.response_data.melt(id_vars=[resp_id_col], value_vars=self.correct_answers.keys(),
                                                var_name='Pair', value_name='Response')
        correctness_long = self.response_data.melt(id_vars=[resp_id_col],
                                                   value_vars=[f'{p}_Correct' for p in self.correct_answers.keys()],
                                                   var_name='Pair_Correct_Col', value_name='Correct')
        correctness_long['Pair'] = correctness_long['Pair_Correct_Col'].str.replace('_Correct', '')
        response_long = response_long.merge(correctness_long[[resp_id_col, 'Pair', 'Correct']],
                                            on=[resp_id_col, 'Pair'])

        q_to_pair = {f'Q{i + 1}': f'Pair{i + 1}' for i in range(6)}
        self.combined_data['Pair'] = self.combined_data['Question'].map(q_to_pair)
        self.combined_data = self.combined_data.merge(response_long, left_on=[et_id_col, 'Pair'],
                                                      right_on=[resp_id_col, 'Pair'], how='left')
        self.combined_data['Answer_Correctness'] = self.combined_data['Correct'].map(
            {True: 'Correct', False: 'Incorrect'})

        # Identify numeric and time columns for later use
        self.numeric_cols = self.combined_data.select_dtypes(include=np.number).columns.tolist()
        self.time_metrics = [c for c in self.numeric_cols if
                             any(k in c.lower() for k in ['time', 'duration', 'fixation'])]
        print("Data loading complete.")
        return self  # Return self for chaining

    def analyze_rq1_metric(self, metric):
        """Analyzes a single metric for RQ1."""
        if metric not in self.combined_data.columns:
            return None, "Metric not found."

        correct = self.combined_data.loc[self.combined_data['Answer_Correctness'] == 'Correct', metric].dropna()
        incorrect = self.combined_data.loc[self.combined_data['Answer_Correctness'] == 'Incorrect', metric].dropna()

        # Perform t-test
        t_stat, p_val = stats.ttest_ind(incorrect, correct, equal_var=False, nan_policy='omit')

        # Create plot
        fig, ax = plt.subplots(figsize=(8, 6))
        sns.boxplot(data=self.combined_data, x='Answer_Correctness', y=metric, ax=ax, palette=['#66b3ff', '#ff9999'])
        ax.set_title(f'Comparison of "{metric}" by Answer Correctness', fontsize=14)
        ax.set_xlabel("Answer Correctness")
        ax.set_ylabel(metric)
        plt.tight_layout()

        # Create summary text
        summary = f"""
        ### Analysis for: **{metric}**
        - **Mean (Correct Answers):** {correct.mean():.4f}
        - **Mean (Incorrect Answers):** {incorrect.mean():.4f}
        - **T-test p-value:** {p_val:.4f}

        **Conclusion:**
        - {'There is a **statistically significant** difference between the groups (p < 0.05).' if p_val < 0.05 else 'There is **no statistically significant** difference between the groups (p >= 0.05).'}
        """
        return fig, summary

    def run_prediction_model(self, test_size, n_estimators):
        """Runs the RandomForest model with given parameters for RQ2."""
        leaky_features = ['Total_Correct', 'Overall_Accuracy', 'Correct']
        features_to_use = [col for col in self.numeric_cols if col not in leaky_features]

        features = self.combined_data[features_to_use].copy()
        target = self.combined_data['Answer_Correctness'].map({'Correct': 1, 'Incorrect': 0})

        valid_indices = target.notna()
        features, target = features[valid_indices], target[valid_indices]
        features = features.fillna(features.median()).fillna(0)

        if len(target.unique()) < 2:
            return "Not enough classes to train the model.", None

        X_train, X_test, y_train, y_test = train_test_split(features, target, test_size=test_size, random_state=42,
                                                            stratify=target)
        scaler = StandardScaler()
        X_train_scaled, X_test_scaled = scaler.fit_transform(X_train), scaler.transform(X_test)

        model = RandomForestClassifier(n_estimators=n_estimators, random_state=42, class_weight='balanced')
        model.fit(X_train_scaled, y_train)
        y_pred_proba = model.predict_proba(X_test_scaled)[:, 1]
        y_pred = model.predict(X_test_scaled)

        # Generate results
        report = classification_report(y_test, y_pred, target_names=['Incorrect', 'Correct'], output_dict=True)
        auc_score = roc_auc_score(y_test, y_pred_proba)

        report_df = pd.DataFrame(report).transpose().round(3)
        report_md = f"""
        ### Model Performance
        - **AUC Score:** **{auc_score:.4f}**
        - **Overall Accuracy:** {report['accuracy']:.3f}

        **Classification Report:**
        {report_df.to_markdown()}
        """

        # Feature importance plot
        feature_importance = pd.DataFrame({'Feature': features.columns, 'Importance': model.feature_importances_})
        feature_importance = feature_importance.sort_values('Importance', ascending=False).head(15)

        fig, ax = plt.subplots(figsize=(10, 8))
        sns.barplot(data=feature_importance, x='Importance', y='Feature', ax=ax, palette='viridis')
        ax.set_title(f'Top 15 Predictive Features (n_estimators={n_estimators})', fontsize=14)
        plt.tight_layout()

        return report_md, fig


# --- DATA SETUP (RUNS ONCE AT STARTUP) ---
def setup_and_load_data():
    """Clones the repo if not present and loads data."""
    repo_url = "https://github.com/RextonRZ/GenAIEyeTrackingCleanedDataset"
    repo_dir = "GenAIEyeTrackingCleanedDataset"

    if not os.path.exists(repo_dir):
        print(f"Cloning data repository from {repo_url}...")
        git.Repo.clone_from(repo_url, repo_dir)
    else:
        print("Data repository already exists.")

    base_path = os.path.join(repo_dir, "cleaned_dataset")
    response_file = os.path.join(repo_dir, "response_sheet", "GenAI_Response_Sheet.xlsx")

    analyzer = EnhancedAIvsRealGazeAnalyzer().load_and_process_data(base_path, response_file)
    return analyzer


print("Starting application setup...")
analyzer = setup_and_load_data()
print("Application setup complete. Ready for interaction.")


# --- GRADIO INTERACTIVE FUNCTIONS ---
def update_rq1_visuals(metric_choice):
    """Called by Gradio when the dropdown for RQ1 changes."""
    if not metric_choice:
        return None, "Please select a metric from the dropdown."
    plot, summary = analyzer.analyze_rq1_metric(metric_choice)
    return plot, summary


def update_rq2_model(test_size, n_estimators):
    """Called by Gradio when sliders for RQ2 change."""
    n_estimators = int(n_estimators)  # Ensure it's an integer
    report, plot = analyzer.run_prediction_model(test_size, n_estimators)
    return report, plot


# --- GRADIO INTERFACE DEFINITION ---
description = """
# Interactive Dashboard: AI vs. Real Gaze Analysis
Explore the eye-tracking dataset by interacting with the controls below. The data is automatically loaded from the public GitHub repository.
"""

with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown(description)

    with gr.Tabs():
        with gr.TabItem("RQ1: Viewing Time vs. Correctness"):
            gr.Markdown("### Does viewing time differ based on whether a participant's answer was correct?")
            with gr.Row():
                with gr.Column(scale=1):
                    rq1_metric_dropdown = gr.Dropdown(
                        choices=analyzer.time_metrics,
                        label="Select a Time-Based Metric to Analyze",
                        value=analyzer.time_metrics[0] if analyzer.time_metrics else None
                    )
                    rq1_summary_output = gr.Markdown(label="Statistical Summary")
                with gr.Column(scale=2):
                    rq1_plot_output = gr.Plot(label="Metric Comparison")

        with gr.TabItem("RQ2: Predicting Correctness from Gaze"):
            gr.Markdown("### Can we build a model to predict answer correctness from gaze patterns?")
            with gr.Row():
                with gr.Column(scale=1):
                    gr.Markdown("#### Tune Model Hyperparameters")
                    rq2_test_size_slider = gr.Slider(
                        minimum=0.1, maximum=0.5, step=0.05, value=0.3, label="Test Set Size"
                    )
                    rq2_estimators_slider = gr.Slider(
                        minimum=10, maximum=200, step=10, value=100, label="Number of Trees (n_estimators)"
                    )
                    rq2_report_output = gr.Markdown(label="Model Performance Report")
                with gr.Column(scale=2):
                    rq2_plot_output = gr.Plot(label="Feature Importance")

    # Wire up the interactive components
    rq1_metric_dropdown.change(
        fn=update_rq1_visuals,
        inputs=[rq1_metric_dropdown],
        outputs=[rq1_plot_output, rq1_summary_output]
    )

    # Use .release to only update when the user lets go of the slider
    rq2_test_size_slider.release(
        fn=update_rq2_model,
        inputs=[rq2_test_size_slider, rq2_estimators_slider],
        outputs=[rq2_report_output, rq2_plot_output]
    )
    rq2_estimators_slider.release(
        fn=update_rq2_model,
        inputs=[rq2_test_size_slider, rq2_estimators_slider],
        outputs=[rq2_report_output, rq2_plot_output]
    )

    # Load initial state
    demo.load(
        fn=update_rq1_visuals,
        inputs=[rq1_metric_dropdown],
        outputs=[rq1_plot_output, rq1_summary_output]
    )
    demo.load(
        fn=update_rq2_model,
        inputs=[rq2_test_size_slider, rq2_estimators_slider],
        outputs=[rq2_report_output, rq2_plot_output]
    )

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