| name | analyze-results |
| description | 分析数模求解结果，计算统计量，生成对比表和洞察。当用户说'分析结果'、'结果对比'、'analyze results'、或需要解读计算结果时使用。 |
| argument-hint | ["results-path-or-description"] |
| allowed-tools | Bash(*), Read, Grep, Glob, Write, Edit, Agent |
分析求解结果
Analyze: $ARGUMENTS
Workflow
Step 1: 定位结果文件
Find all relevant result files:
- Check
results/, output/, figures/, or project-specific output directories
- Parse JSON/CSV/Excel results into structured data
- Read solver logs for convergence information
Step 2: 构建对比表
Organize results by:
- 子问题: which sub-problem each result belongs to
- 方法: which mathematical method was used
- 关键指标: R², RMSE, 目标函数值, 残差, 误差率 etc.
- 与预期对比: compare against expected values, theoretical bounds, or baseline methods
Step 3: 统计分析
- If multiple runs: report mean +/- std, check robustness
- If sweeping a parameter: identify sensitivity trends (monotonic, U-shaped, plateau)
- Perform error analysis: residual distribution, prediction vs actual
- Check model fit: R², adjusted R², F-statistic, p-values
- Flag outliers or suspicious results (unreasonable values, non-convergence)
Step 4: 生成洞察
For each finding, structure as:
- 观察: what the data shows (with numbers)
- 解读: why this result makes sense (or doesn't)
- 意义: what this means for the competition problem
- 下一步: what analysis or improvement would strengthen the result
Step 5: 生成图表建议
Based on the analysis, suggest figures for the paper:
- 拟合曲线 (predicted vs actual)
- 残差分布图
- 灵敏度分析图 (parameter sensitivity)
- 误差随参数变化图
- 结果对比柱状图/表格
Step 6: 更新文档
If findings are significant:
- Propose updates to MODEL_REPORT.md or SOLVE_PLAN.md
- Draft a concise finding statement (1-2 sentences)
- Save analysis to
RESULTS_ANALYSIS.md
Output Format
Always include:
- Raw data table (数值结果)
- Key findings (numbered, concise)
- Statistical validation (误差分析、模型检验)
- Suggested figures for paper
- Suggested next steps (if any)
