def __plot_top_k_drifts()

in src/drift_visualizer.py [0:0]

• 12 lines of code
• 3 McCabe index (conditional complexity)

def __plot_top_k_drifts(drift_df, feature, ylabel, top_k):
    df = drift_df.copy()
    
    feature_score = [(feat, max(df[df.Feature == feat][feature]))
                     for feat in df.Feature.value_counts().keys()]
    
    # Sort by drift score to pick top k features
    feature_score = [x[0] for x in sorted(feature_score, key=lambda x: x[1], reverse=True)][:top_k]
    
    df = df[['Feature', feature, 'Time']]
    df = df[df.Feature.isin(feature_score)]
    df = df.pivot(index='Time', columns='Feature', values=feature)

    ax = df.plot()
    ax.set_xlabel('Time')
    ax.set_ylabel(ylabel)
    ax.legend(bbox_to_anchor=(1.1, 1.05))