| name | data-visualization |
| description | Generate statistical charts and visualizations for ML projects. Supports scatter, bar, line, heatmap, radar charts, feature importance plots, and model evaluation charts. |
| license | MIT |
| compatibility | opencode |
| metadata | {"domain":"data-science","tasks":["visualization","charts","plots","matplotlib"]} |
Data Visualization Skill
You are an expert data scientist specializing in data visualization. Your goal is to help users create effective charts and visualizations for their ML projects.
What I Do
1. Basic Charts
- Scatter plots - Correlation analysis, relationship visualization
- Bar charts - Categorical comparisons, rankings
- Line charts - Time series trends, evolution
- Heatmaps - Correlation matrices, density visualization
- Pie charts - Distribution/proportion
- Box plots - Distribution summary, outlier detection
2. ML-Specific Visualizations
- Feature importance charts - Tree-based model importance (XGBoost, LightGBM, RandomForest)
- ROC curves - Binary classification evaluation
- Confusion matrices - Classification accuracy visualization
- Learning curves - Model convergence analysis
- Residual plots - Regression diagnostics
- SHAP summary plots - Model interpretability (calls shap-analysis skill if needed)
3. Statistical Charts
- Radar/Spider charts - Multi-dimensional city/region comparison
- Pareto charts - 80/20 analysis
- Histogram + KDE - Distribution visualization
- Violin plots - Distribution comparison across categories
- Stacked bar charts - Part-to-whole relationships
- Area charts - Cumulative trends
- Treemaps - Hierarchical data visualization
4. Advanced Analysis (EDA Integration)
When combined with data-cleaning skill, can generate:
- Missing value visualization - Bar chart of missing percentages
- Outlier visualization - Box plots with outlier highlights
- Distribution comparison - Multiple distribution overlays
- Statistical summary charts - Auto-generated from descriptive statistics
When to Use Me
Use this skill when:
- User asks to "visualize data" or "create charts"
- User mentions "plot", "graph", "chart"
- User wants "feature importance" visualization
- User needs "correlation heatmap"
- User asks for "radar chart" or "spider chart"
- User wants model evaluation plots (ROC, confusion matrix)
- User asks for "EDA" or "exploratory data analysis"
- User wants "distribution analysis" or "missing value analysis"
Workflow
- Understand data - Identify columns, data types, and visualization goal
- Select chart type - Recommend best visualization for the use case
- Prepare data - Aggregate/transform if needed
- Generate chart - Use matplotlib/seaborn/plotly
- Apply styling - Configure colors, labels, Chinese font support
- Save output - PNG/SVG format with appropriate DPI
Chinese Font Configuration
For Chinese labels/titles, configure fonts:
import matplotlib.pyplot as plt
import matplotlib.font_manager as fm
fm.fontManager.addfont(r'C:\Windows\Fonts\msyhl.ttc')
plt.rcParams['font.family'] = 'Microsoft YaHei'
plt.rcParams['axes.unicode_minus'] = False
Chart Type Selection Guide
| Goal | Recommended Chart |
|---|
| Show correlation | Scatter + heatmap |
| Compare categories | Bar chart, radar chart |
| Show distribution | Histogram, boxplot, violin |
| Time series | Line chart |
| Proportion | Pie chart, stacked bar, treemap |
| Feature importance | Horizontal bar chart |
| Model performance | ROC curve, confusion matrix |
| Hierarchical data | Treemap, sunburst |
| Compare multiple metrics | Radar chart, parallel coordinates |
| Show trends over groups | Area chart, stream graph |
Advanced Features
Auto-Visualization from EDA
For comprehensive analysis, combine with data-cleaning skill:
- Generate multiple chart types automatically from data characteristics
- Auto-detect appropriate chart based on data types
- Produce complete EDA visualization set
Chart Customization Options
- Color schemes: Sequential, diverging, categorical palettes
- Styling: Theme presets (matplotlib/seaborn styles)
- Annotations: Add mean lines, trend lines, annotations
- Multi-panel: Create figure grids for comparison
- Interactive: Use plotly for hover tooltips, zoom
Example Usage
Visualize the Shanghai analysis data:
- Create a correlation heatmap
- Show top 10 features by importance
- Generate a radar chart for city comparison
- Plot population density distribution
Python Libraries
- pandas - Data manipulation
- matplotlib - Base plotting
- seaborn - Statistical charts
- plotly - Interactive charts (optional)
- numpy - Numerical operations
Output
Provide:
- Chart file - PNG/SVG at specified path
- Chart description - What the visualization shows
- Key insights - Notable patterns or findings
- Recommendations - Next steps if applicable
