eda

Prompt file imported from taimo3810/kaggle-with-ai-template (.claude/commands/eda.md). Fill in {{arguments}} before use. Copyright stays with the author.

Exploratory Data Analysis (EDA)

Perform EDA on the following data file using Codex CLI for intelligent analysis.

Target Data: {{arguments}}

Context Information

Refer to the contents of the following files to understand data meaning and evaluation metrics before analyzing.

• @COMPETITION.md
• @DATASET.md
• @METRIC.md

Data Directory Structure

• data/raw/: Original immutable data
• data/processed/: Processed data features

Instructions

1. Use Codex for Data Analysis

Execute Codex CLI to perform intelligent EDA:

codex exec --full-auto --sandbox read-only --cd /Users/taikimori/kaggle-with-ai-template \
  "以下のデータファイルに対してEDA（探索的データ分析）を実行してください: {{arguments}}

  分析内容:
  1. データの基本統計量 (describe(), info())
  2. 欠損値の確認と可視化
  3. ターゲット変数の分布
  4. 主要特徴量とターゲットの相関
  5. 画像データの場合はサンプル画像の表示
  6. 座標データの場合は地理的分布の可視化

  結果をnotebook/eda.ipynbに保存してください。"

2. Alternative: Manual Codex Commands

For specific analyses, use targeted Codex commands:

Basic Statistics

codex exec --full-auto --sandbox read-only --cd /Users/taikimori/kaggle-with-ai-template \
  "data/raw/のCSVファイルを読み込んで基本統計量を出力してください"

Missing Values Analysis

codex exec --full-auto --sandbox read-only --cd /Users/taikimori/kaggle-with-ai-template \
  "データの欠損値を分析して可視化してください"

Geographic Distribution (for location data)

codex exec --full-auto --sandbox read-only --cd /Users/taikimori/kaggle-with-ai-template \
  "緯度経度データの地理的分布を可視化してください"

Image Sample Display (for image data)

codex exec --full-auto --sandbox read-only --cd /Users/taikimori/kaggle-with-ai-template \
  "images/ディレクトリからサンプル画像を10枚表示してください"

3. Create Notebook

After Codex analysis, ensure the notebook is saved:

codex exec --full-auto --sandbox read-only --cd /Users/taikimori/kaggle-with-ai-template \
  "これまでの分析をnotebook/eda.ipynbにJupyter Notebook形式でまとめてください。
  - Markdownセルで分析意図と結果の解釈を記述
  - matplotlib/seabornで可視化
  - 英語で記述"

Codex Parameters Reference

Expected EDA Output

For Tabular Data

• Shape and data types
• Missing value counts and percentages
• Statistical summaries (mean, std, min, max, quartiles)
• Target variable distribution
• Feature correlations heatmap

For Image Data (atmaCup #23)

• Image count and file sizes
• Sample image grid display
• Image dimension verification
• Mask overlay visualization
• Camera angle distribution

For Geographic Data

• Coordinate range and distribution
• Map visualization with scatter plot
• Density heatmap
• Scene clustering visualization

Notes

• Codex runs in read-only sandbox mode for safe analysis
• If data size is large, Codex will automatically consider sampling
• Results are saved to notebook/eda.ipynb
• For complex visualizations, Codex may generate multiple cells