Skip to content

README.md

Latest commit

 

History

History
81 lines (64 loc) · 3.01 KB

README.md

File metadata and controls

81 lines (64 loc) · 3.01 KB
Raw

TW Loan Acquisition Chatbot PART 1

TVS Loan Recommendation Model

This model provides a customer-friendly chatbot taking minimum questions from users, processes using ML pipeline and provides an instant loan eligibility decision (approve/decline).

By further integrating the conversational UI with a personalized recommendation layer, where instead of rejecting a customer outright for the specific model, it evaluates all models under the given Make_Code and returns the list of approved/eligible options for the user.

This approach softens the impact of rejection, gives user multiple financing choices instantly, improves satisfaction & increases likelihood of conversion.

Basic Overview

  1. Problem Space : There exists fundamental tension between 2 critical objectives - maximizing efficiency of loan application process & minimizing credit risk carried by new applicants.
  2. Proposed Solution Diagram :

Detailed Overview

  1. User Input (9 Fields)

    • Age
    • Gender
    • Pincode
    • Qualifications
    • Employment_Type
    • Net_salary
    • Make_Code
    • Loan_Amount
    • PAST_LOANS_ACTIVE

  2. Lookup Table

    • PINCODE_MAP : Dictionary grouped by Pincode with State, State_Avg_Salary, Final_Tier (14031 entries)
    • MAKE_CODE_MAP : Dictionary of Make_Code mapped to Model_Description, Product_Code, Avg_Model_Price (13 entries)

  3. Feature Engineering

    • Derived Features
      • Avg Model Price
      • Income Presence (Income / No Income / Guarantor)
    • Binning
      • Age Group
      • LTV Band (Low / Medium / High)
      • Income Tier (Low → Very High)
      • Loan Amount Band
    • Relational Features
      • Employment × Gender
      • Income-to-Loan Ratio
      • Age-to-Loan Ratio
      • Salary vs State Average
      • Past Loans × Employment
      • State-Pincode Zone

  4. Model Used : LightGBM (Gradient Boosted Decision Trees)

  5. Model Performance

Metrics Value
Accuracy 0.9431
Precision 0.9458
Recall 0.9970
F1 Score 0.9707
ROC AUC Score 0.7651 (more critical to separate safe borrowers from risky ones)
KS Statistics 0.4 (in credit risk modeling, KS = 0.40 puts model in strong performance category)
Processing Time ~0.04 seconds (reduced with real time feature engineering & table creation)
Chat Session Time ~26.63 seconds (dependent on user)

Risks & Trade-offs

Risk Description
False Positives Higher NPAs (approving risky customers)
False Negatives Losing eligible customers

Conclusion

This solution demonstrates how :

  • Conversational AI + ML + Recommendation Systems can transform lending
  • Real-time eligibility can directly guide purchase decisions
  • Integrating approval with bike recommendations improves both UX and business outcomes

It effectively balances : Speed (customer experience), Accuracy (risk control) & Personalization (bike recommendations)