EricCRX/books-tabular-dataset

---

language: en tags: - tabular - education - books task_categories: - tabular-classification

📄 Model Card: Books Tabular Dataset

1. Purpose

This dataset was created for educational purposes in the context of Homework 1 (Dealing with Data). The goal is to provide a small but structured tabular dataset that allows students to practice working with real-world features, preprocessing, augmentation, and uploading to Hugging Face. The dataset supports tasks such as classification, exploratory data analysis (EDA), and simple modeling.

---

2. Composition

• Domain: Books (physical objects collected by the student).
• Samples: 30 unique real-world books.
• Features (≥5):
  • Length_cm (continuous, cm)
  • Width_cm (continuous, cm)
  • Thickness_cm (continuous, cm)
  • Pages (integer)
  • Hardcover (categorical: Yes/No)
  • Cover_Color (categorical: dominant color)
• Target (binary):
  • Is_Textbook = {Yes, No}

---

3. Collection Process

• Source: All samples were collected manually from physical books available to the student.
• Method: Each book was measured with a ruler (length, width, thickness) and checked for number of pages. Hardcover status and cover color were recorded by visual inspection.
• Uniqueness: Each of the 30 samples corresponds to a different book; no duplicates are present.

---

4. Preprocessing & Augmentation

• Preprocessing:
  • Units standardized to centimeters for dimensions.
  • Pages recorded as integers.
  • Colors normalized to a small set of categorical labels (e.g., red, blue, green).
• Augmentation:
  • Synthetic data (≥300 rows) was generated by applying label-preserving transformations, such as:
    • Adding Gaussian noise to continuous features.
    • Slight perturbations to page counts.
    • Resampling categorical features with the same distribution.
  • Augmented samples maintain the same schema as the original split.

---

5. Labels

• Target Variable: Is_Textbook
• Type: Binary classification
• Distribution: Labels reflect the real-world sample distribution; balance was not artificially enforced.

---

6. Splits

• Original split: 30 manually measured samples (original)
• Augmented split: 300 synthetically generated samples (augmented)

---

7. Exploratory Data Analysis (EDA)

Summary statistics for the 30 original samples:

Feature	Mean	Std	Min	Max
Length_cm	~21.3	~3.2	15.0	28.0
Width_cm	~14.6	~2.5	10.0	19.0
Thickness_cm	~2.4	~1.1	0.8	5.0
Pages	~280	~110	100	520

Exploratory Data Analysis (EDA)

Here are some plots from the dataset:

---

8. Intended Use / Limitations

• Use: Teaching and practice with tabular data, preprocessing, augmentation, and Hugging Face uploads.
• Limitations:
  • Small dataset, not suitable for production ML.
  • Augmented data may not match true real-world distributions.
  • Cover color is subjective.

---

9. Ethical Considerations

• No sensitive or personally identifiable information (PII).
• Objects are inanimate (books).
• No foreseeable ethical risks.

---

10. License

This dataset is released under the MIT License, permitting educational and research use with attribution.

---

11. AI Usage Disclosure

• Original 30 samples: Collected and measured manually.
• Synthetic augmentation & Model Card drafting: Supported by generative AI (ChatGPT).
• Human verification: Final dataset and documentation checked manually.