Parallel coordinates plot with test score above 0.85

Module 3. Hyperparameter tuning
#1

How can i make sure that i select test score of 0.85 and above? the plot doesn’t show the selected parameter or metric ranges.

image
image974×596 209 KB

#2

There is no strict way to find it. When we state above 0.85 is to mention to top best performing model and give an idea of where to place the cursor. With your current selection, it should already be good to answer the subsequent questions.

1 Like
#3

Thank you @glemaitre58. Indeed that’s what i did. Just wondered if there is a another way to do it that i am now aware of. Thank you for confirming this!

#4

I used plotly graph_objects with tickvals to display 8.5 tick with below code:

import plotly.graph_objects as go

fig = go.Figure(data=
    go.Parcoords(
        line = dict(color = cv_results["mean_test_score"]),
        dimensions = list([
            dict(
                label = 'Learning rate', values = cv_results['param_classifier__learning_rate']),
            dict(
                label = 'Max leaf nodes', values = cv_results['param_classifier__max_leaf_nodes']),
            dict(
                label = 'Max bins', values = cv_results['param_classifier__max_bins']),
            dict(
                label = 'Min samples Leaf', values = cv_results['param_classifier__min_samples_leaf']),
            dict(
                label = 'L2 regularization', values = cv_results['param_classifier__l2_regularization']),
            dict(
                tickvals = [0.1,0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.85, 0.9],
                label = 'Mean test score', values = cv_results['mean_test_score']),
            
        ])
    )
)

fig.update_layout(
    plot_bgcolor = 'white',
    paper_bgcolor = 'white'
)

fig.show()

image
image1163×659 234 KB

#5

This is a good programmatic solution. You might need to add the log2/log10 space to get the same axis as previously if needed.