在绘图中对混淆矩阵进行归一化的自定义颜色

Question

我正在尝试创建一个混淆矩阵，其中

• 中性值为 0，应显示为白色
• 正值应显示为绿色，数字越大，绿色越多。最接近 0，绿色较少（与白色混合）
• 负值应显示为红色，数字越小，红色越多。最接近 0，红色较少（与白色混合）

我希望我的最高腹肌数能够正常化。这意味着，最高的绝对数决定纯绿色/红色。

import plotly.graph_objects as go
import numpy as np

cf_matrix = np.array([[395, -5], [-100, 20]])

group_names = ["TP", "FP", "FN", "TN"]
group_values = [395, -5, -100, 20]


labels = [f"{v1}\n{v2}" for v1, v2 in zip(group_names, group_values)]
labels = np.asarray(labels).reshape(2, 2)

# Define the colors for 0, positive, and negative values
zero_color = 'white'
positive_color = 'green'
negative_color = 'red'

# Create a custom color scale
colors = [
    (0, negative_color),
    (0.5 / np.max(abs(custom_cf_matrix)), zero_color),  # Normalize the midpoint value to 0.5
    (1, positive_color)
]

# Create the figure
fig = px.imshow(
    custom_cf_matrix,
    labels={"x": "Predicted Label", "y": "True Label"},
    color_continuous_scale=colors,
    range_color=[np.max(abs(custom_cf_matrix)), np.max(abs(custom_cf_matrix))],
    width=500,
    height=500,
)

fig.update_xaxes(side="bottom")
fig.update_yaxes(side="left")

# Update the annotations to use black font color
annotations = [
    dict(
        text=text,
        x=col,
        y=row,
        font=dict(color="black", size=16),  # Set font color to black
        showarrow=False,
        xanchor="center",
        yanchor="middle",
    )
    for row in range(2)
    for col, text in zip(range(2), labels[row])
]

fig.update_layout(
    title="Value-Weighted Confusion Matrix",
    title_x=0.25,  # Center the title horizontally
    annotations=annotations,
)

fig.update_xaxes(tickvals=[0, 1], ticktext=["0", "1"], showticklabels=True)
fig.update_yaxes(tickvals=[0, 1], ticktext=["0", "1"], showticklabels=True)

# Display the plot
st.plotly_chart(fig)


fig.show()