Skip to content

Commit 72c7b04

Browse files
github-actions[bot]github-actions[bot]
authored
feat(altair): implement shap-summary (#2987)
## Implementation: `shap-summary` - altair Implements the **altair** version of `shap-summary`. **File:** `plots/shap-summary/implementations/altair.py` **Parent Issue:** #2923 --- :robot: *[impl-generate workflow](https://github.com/MarkusNeusinger/pyplots/actions/runs/20612948253)* --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
1 parent da1c883 commit 72c7b04

2 files changed

Lines changed: 136 additions & 0 deletions

File tree

‎plots/shap-summary/implementations/altair.py‎

Lines changed: 108 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,108 @@
1+
""" pyplots.ai
2+
shap-summary: SHAP Summary Plot
3+
Library: altair 6.0.0 | Python 3.13.11
4+
Quality: 92/100 | Created: 2025-12-31
5+
"""
6+
7+
import altair as alt
8+
import numpy as np
9+
import pandas as pd
10+
11+
12+
# Generate synthetic SHAP values for a model explanation visualization
13+
np.random.seed(42)
14+
n_samples = 300
15+
n_features = 10
16+
17+
# Feature names representing typical ML model inputs
18+
feature_names = [
19+
"Account Age (months)",
20+
"Transaction Count",
21+
"Avg Transaction ($)",
22+
"Credit Score",
23+
"Income ($K)",
24+
"Debt Ratio",
25+
"Payment History",
26+
"Account Balance ($)",
27+
"Login Frequency",
28+
"Support Tickets",
29+
]
30+
31+
# Create synthetic feature values (normalized to 0-1 for color mapping)
32+
feature_values = np.random.rand(n_samples, n_features)
33+
34+
# Create synthetic SHAP values with varying importances per feature
35+
# Higher importance = wider spread of SHAP values
36+
feature_importances = np.array([0.25, 0.20, 0.15, 0.12, 0.10, 0.07, 0.05, 0.03, 0.02, 0.01])
37+
shap_values = np.zeros((n_samples, n_features))
38+
39+
for i in range(n_features):
40+
# Create SHAP-like values: correlation with feature value + noise
41+
# More important features have larger SHAP value spreads
42+
base_effect = (feature_values[:, i] - 0.5) * feature_importances[i] * 4
43+
noise = np.random.randn(n_samples) * feature_importances[i] * 0.5
44+
shap_values[:, i] = base_effect + noise
45+
46+
# Sort features by importance (already ordered, but calculate mean abs for verification)
47+
mean_abs_shap = np.mean(np.abs(shap_values), axis=0)
48+
feature_order = np.argsort(mean_abs_shap)[::-1]
49+
feature_order_names = [feature_names[i] for i in feature_order]
50+
51+
# Build dataframe for Altair
52+
rows = []
53+
for feat_idx in feature_order:
54+
for sample_idx in range(n_samples):
55+
rows.append(
56+
{
57+
"Feature": feature_names[feat_idx],
58+
"SHAP Value": shap_values[sample_idx, feat_idx],
59+
"Feature Value": feature_values[sample_idx, feat_idx],
60+
"Importance": mean_abs_shap[feat_idx],
61+
}
62+
)
63+
64+
df = pd.DataFrame(rows)
65+
66+
# Create the SHAP summary plot
67+
scatter = (
68+
alt.Chart(df)
69+
.mark_circle(opacity=0.7, stroke="#333333", strokeWidth=0.3)
70+
.encode(
71+
x=alt.X(
72+
"SHAP Value:Q",
73+
title="SHAP Value (Impact on Model Output)",
74+
axis=alt.Axis(titleFontSize=22, labelFontSize=18, grid=True, gridOpacity=0.3),
75+
),
76+
y=alt.Y("Feature:N", title=None, sort=feature_order_names, axis=alt.Axis(labelFontSize=18)),
77+
color=alt.Color(
78+
"Feature Value:Q",
79+
scale=alt.Scale(scheme="blueorange", domain=[0, 1]),
80+
legend=alt.Legend(
81+
title="Feature Value", titleFontSize=18, labelFontSize=16, orient="right", gradientLength=200
82+
),
83+
),
84+
size=alt.value(80),
85+
yOffset=alt.YOffset("jitter:Q", scale=alt.Scale(domain=[-1, 1], range=[-15, 15])),
86+
)
87+
.transform_calculate(jitter="random() * 2 - 1")
88+
)
89+
90+
# Add vertical line at x=0
91+
zero_line = (
92+
alt.Chart(pd.DataFrame({"x": [0]})).mark_rule(color="#333333", strokeWidth=2, strokeDash=[5, 3]).encode(x="x:Q")
93+
)
94+
95+
# Combine scatter and zero line
96+
chart = (
97+
(zero_line + scatter)
98+
.properties(
99+
width=1400, height=850, title=alt.Title("shap-summary · altair · pyplots.ai", fontSize=28, anchor="middle")
100+
)
101+
.configure_axis(labelFontSize=18, titleFontSize=22)
102+
.configure_view(strokeWidth=0)
103+
.configure_legend(titleFontSize=18, labelFontSize=16)
104+
)
105+
106+
# Save outputs
107+
chart.save("plot.png", scale_factor=3.0)
108+
chart.save("plot.html")

‎plots/shap-summary/metadata/altair.yaml‎

Lines changed: 28 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,28 @@
1+
library: altair
2+
specification_id: shap-summary
3+
created: '2025-12-31T05:53:37Z'
4+
updated: '2025-12-31T06:01:03Z'
5+
generated_by: claude-opus-4-5-20251101
6+
workflow_run: 20612948253
7+
issue: 2923
8+
python_version: 3.13.11
9+
library_version: 6.0.0
10+
preview_url: https://storage.googleapis.com/pyplots-images/plots/shap-summary/altair/plot.png
11+
preview_thumb: https://storage.googleapis.com/pyplots-images/plots/shap-summary/altair/plot_thumb.png
12+
preview_html: https://storage.googleapis.com/pyplots-images/plots/shap-summary/altair/plot.html
13+
quality_score: 92
14+
review:
15+
strengths:
16+
- Excellent implementation of SHAP summary plot matching specification requirements
17+
- Effective use of Altair declarative syntax with proper encoding types (Q for quantitative,
18+
N for nominal)
19+
- Good jittering implementation using transform_calculate for random vertical offsets
20+
- Clean code structure following KISS principles
21+
- Appropriate color scheme (blueorange) matching SHAP convention of low=blue, high=red/orange
22+
- Features properly sorted by mean absolute SHAP value (importance)
23+
weaknesses:
24+
- Legend positioning could be closer to the plot area to reduce isolation
25+
- Grid styling is minimal - could add subtle horizontal grid lines to help trace
26+
feature rows
27+
- Interactive features not utilized (no tooltips or zoom/pan) which is an Altair
28+
strength

0 commit comments

Comments
 (0)