Part 4: Downloading the Dataset

4. Why Do We Need a Dataset?

Before training models, we must understand the role of data in Machine Learning.

A machine learning model has no innate human intelligence. It cannot guess whether a customer is reliable or risky based on intuition. Instead, it relies entirely on historical evidence.

A dataset is a structured compilation of past historical records. In our case, the dataset contains details of 4,269 previous loan applications, each labeled with its final outcome (Approved or Rejected). The model acts like a student looking at past exams: it analyzes these historical records, learns which traits (such as high incomes and strong credit scores) correlate with approvals, and uses those patterns to evaluate new, unseen applicants.

In machine learning, we say “Garbage In, Garbage Out”. The accuracy and fairness of your model depend entirely on the quality and size of your dataset. If your dataset contains incorrect records or represents only one type of applicant, your model will make poor decisions in production.

---

4.1 The Loan Dataset

For this project, we utilize the official Loan Approval Prediction Dataset on Kaggle.

Place the downloaded loan_approval_dataset.csv file inside your project’s data folder. The final path must be: smart_lender/data/loan_approval_dataset.csv

---

4.2 Dataset Columns Explained

The raw fields inside loan_approval_dataset.csv are structured as follows:

• Applicant Demographics:
  • loan_id: A unique numerical identifier for each application.
  • no_of_dependents: The number of family members dependent on the applicant.
  • education: The educational qualification status (Graduate or Not Graduate).
  • self_employed: Whether the applicant is self-employed (Yes or No).
• Financial & Loan Details:
  • income_annum: The annual income of the applicant.
  • loan_amount: The total loan amount requested by the applicant.
  • loan_term: The repayment period requested for the loan (in years).
• Credit Worthiness & Asset Valuations:
  • cibil_score: The credit score of the applicant (ranging from 300 to 900). A higher score represents lower default risk.
  • residential_assets_value: The valuation of the applicant’s residential property.
  • commercial_assets_value: The valuation of the applicant’s commercial property.
  • luxury_assets_value: The valuation of luxury holdings (such as cars, jewelry).
  • bank_asset_value: Total savings and liquid assets held in bank deposits.
• Target Classification Label:
  • loan_status: The outcome of the application (Approved or Rejected).

---