Natnael-Bacha/icog-labs-naive-bayes-implementation

"# icog-labs-naive-bayes-implementation" # Customer Support Ticket Router A customer-support intent classification system built from scratch using **Multinomial Naive Bayes and NumPy**. The system takes a customer's message and predicts the most appropriate support intent, such as `cancel_order`, `track_order`, `payment_issue`, `recover_password`, `change_order`, `check_invoice`, and `get_refund`. The project also includes a Tkinter GUI for testing individual customer requests and an evaluation system for measuring model performance. --- ## Features * Multinomial Naive Bayes classifier implemented from scratch * NumPy-based model calculations * Text preprocessing * Unigram and bigram features * Stopword removal * Document-frequency based vocabulary filtering * Laplace smoothing * Stratified train/test splitting * Accuracy, Precision, Recall, and F1 evaluation * Classification report * Error analysis * Tkinter graphical interface * No machine-learning library is used for the classifier --- ## Dataset This project uses the **Bitext Customer Support Training Dataset**, specifically: `Bitext_Sample_Customer_Support_Training_Dataset_27K_responses-v11` Credit goes to the dataset creators for providing the customer-support training data used in this project. After duplicate removal, the dataset contained: Original records: 26,872 Duplicates removed: 2,237 Remaining records: 24,635 The dataset contains customer-support intents covering areas such as: * Order cancellation * Order changes * Shipping address changes * Payment issues * Payment methods * Invoice requests * Refunds * Account creation * Account deletion * Password recovery * Delivery information * Order tracking * Customer service * Complaints * Reviews * Newsletter subscriptions ## Technologies * **Python** * **NumPy** — model calculations and numerical operations * **Pandas** — dataset handling * **Tkinter** — graphical user interface --- ## Project Structure ```text customerSupp …