SuperAPI AI Directory

2analysis-modeling

Coding & development

View repository

Updated 10-15 23:32

About this skill

Problem Analysis and Modeling Design

Mathematical Modeling Standards Reference

If domain judgment is needed, read the sections ../_references/math_modeling_norms.md "Problem Understanding and Sub-problem Identification", "Assumptions and Model Building", and "Quick Reference for Common Pitfalls by Problem Type". This file serves only as a standards knowledge base and does not replace the analysis report structure of this stage.

Required Deliverables

In the current working directory's reports/ subdirectory, create or update:

  • reports/ANALYSIS_MODELING_REPORT.md:
    • Problem analysis, sub-problem decomposition, data and attachment understanding, evaluation criteria, key ambiguities, and assumption pre-check.
    • Variables, symbols, model assumptions, objective functions, constraints, solution algorithms, implementation specifications for each sub-problem, and task list for the code stage

Do not write the paper body at this stage, do not generate the final paper/, and do not move figure and table layout tasks forward to this stage.

Workflow

Step 1: Sub-problem Decomposition

Only treat the explicitly numbered top-level problems in the problem statement as sub-problems, such as "Problem 1/2/3" or "问题一/二/三". Do not mistake sub-questions, background descriptions, data descriptions, or submission requirements for independent sub-problems.

At the beginning of ANALYSIS_MODELING_REPORT.md , clearly write:

Adjust the number of problems dynamically according to the problem statement

This problem has X sub-problems in total.

Each sub-problem must state:

  • Input data and known conditions.
  • Decision variables or prediction targets.
  • Objective functions or evaluation metrics.
  • Constraints.
  • Dependencies with other sub-problems.
  • Which figures or tables to draw to present the results.

Step 2: Assumption Sensitivity Pre-check

List key ambiguities and do not rush to settle on a model. For ambiguities that affect the results, provide at least two interpretations and choose by simple verification or logical progression.

Must include in ANALYSIS_MODELING_REPORT.md :

## 假设敏感性预检

### 模糊表述及解释
...

### 快速验算与递进性检查
...

### 最终采用的解释
...

### 绘制的图像和对比表格


If an assumption would make the newly added conditions of subsequent problems have no marginal effect, adjust the interpretation accordingly.

Step 3: Data Understanding and Modeling Route

Perform data understanding for each attachment:

  • Row and column scale and field explanations.
  • Missing values, anomalies, duplicates, and unit inconsistencies.
  • Variables that can be directly used for modeling.
  • Indicators that need to be derived.

Then provide the overall route:

题面 -> 数据清洗(EDA) -> 子问题一模型 -> 子问题二模型 -> 。。。。 -> 结果检验 -> 论文展示

Step 4: Modeling Report

Write in ANALYSIS_MODELING_REPORT.md a complete plan that can be handed to the code stage for implementation.

Each sub-problem must include at least:

  • Problem objective.
  • Symbols and variables.
  • Model assumptions.
  • Objective function.
  • Constraints.
  • Solution method.
  • Inputs and outputs.
  • Key points for code implementation.
  • Result validation methods.

Formulas must be clear enough to be directly implemented at the code stage. Algorithm descriptions must include core steps, stopping conditions, and complexity or feasibility explanations.

Recommended structure:

# 建模报告

## 1. 总体建模框架
## 2. 数据处理方案
## 3. 符号说明
## 4. 问题一模型
## 5. 问题二模型
## 6. 问题三模型
    .... 
## 7. 灵敏度分析与检验方案
## 8. 代码实现任务清单

If the number of sub-problems is not 3, adjust the chapters according to the actual problem statement; do not force it.

Step 5: Interface for the Code Stage

At ANALYSIS_MODELING_REPORT.md write a "Code Implementation Task List" at the end, in the following format:

## 代码实现任务清单

| 任务 | 输入 | 输出 | 方法 | 校验 |
| --- | --- | --- | --- | --- |
| 问题一 | ... | ... | ... | ... |
| 问题二 | ... | ... | ... | ... |

Quality Requirements

  • All conclusions can be traced back to the problem statement or the data.
  • Do not fabricate data fields or values.
  • Do not skip ambiguity analysis.
  • The model must have both a mathematical formulation and be implementable in code.
  • If the data is insufficient or the problem statement is unclear, clearly record the risks and alternative solutions.