Kaiburr Assessment 2025 — Task 5: Consumer Complaint Classification

Overview

This repository contains the solution for Task 5, focused on multi-class text classification of US consumer finance complaints. The code builds an end-to-end machine learning pipeline for pre-processing, feature extraction, model training, evaluation, and prediction using the consumer_complaints.csv dataset from Kaggle.

---

Steps

1. Dataset Download

   • Downloaded via Kaggle CLI: kaggle/us-consumer-finance-complaints
   • Contains open US consumer complaint narrative texts and product category labels.

2. Dependencies

   • Install via:

     pip install pandas scikit-learn matplotlib nltk seaborn
     

3. Code

   • Complete pipeline in consumer_complaint_classification.py
   • Key steps: data cleaning, TF-IDF extraction, four ML models (LR, SVM, RF, NB), evaluation.

4. How to Run

   • Place consumer_complaints.csv and the script in the project root.
   • Run:

     python consumer_complaint_classification.py
     

---

Results

Below are the saved results and evaluation visualizations, each included as PNG from the screenshots/ folder. Every screenshot contains system date/time and my username in the window for verification.

---

1. Model Comparison Table

• Model metrics (accuracy, precision, recall, F1) for all classifiers as output by the classification report.

---

2. Confusion Matrices by Model

Confusion matrices for each classifier, allowing visual inspection of prediction breakdown for each product category:

• Logistic Regression:

• Naive Bayes:

• Random Forest:

• SVM:

---

3. Best Performing Model Proof

Sample output printout and screenshot of the best-performing model selection, along with its prediction evidence.

-

---

Key Files

• consumer_complaint_classification.py — Complete ML pipeline with all steps.
• consumer_complaints.csv — Dataset from Kaggle.

---

Author

Final Year B.Tech CCE
Shyam Anand
October 2025

---

License

This project submitted for Kaiburr Assessment 2025.