Tech Stack

Area	Technologies Used
Language	Python 3.10
Backend API	FastAPI
Machine Learning	Scikit-Learn (TF-IDF + LinearSVC)
Frontend	HTML, CSS, JavaScript
Database	SQLite
Tools	Uvicorn, Pandas, Joblib
Version Control	Git & GitHub

Smart Budgeting Agent — Capstone Project (ITAI-2277)

Author: Derrick K. Gibbs-McGlaston
Instructor: Sitaram Ayyagari
Course: Artificial Intelligence Programming — Capstone
Semester: Fall 2025

Project Overview

The Smart Budgeting Agent is a full-stack AI application that automatically categorizes financial transactions using a machine learning model trained on real-world spending patterns. This project demonstrates mastery of:

• Machine learning model development
• Data preprocessing and feature engineering
• API design with FastAPI
• Frontend integration with JavaScript
• Database persistence using SQLite
• Full-stack AI system architecture and deployment structure
  This repository includes the complete, installable project ready for evaluation and demonstration.

Repository Structure

Capstone_Project/
│
├── backend/ # FastAPI backend + ML model logic
│ ├── app.py # API routes (predict, transactions)
│ ├── db.py # SQLite database handling
│ ├── schema.sql # Database schema
│ ├── train_pipeline.py # ML training script
│ ├── model_pipeline.pkl # Final trained ML model
│ ├── decision_tree_expenses.pkl
│ ├── expenses_preprocessed.csv
│ ├── metrics/
│ └── app.db
│
├── frontend/
│ ├── index.html
│ ├── app.js
│ └── styles.css
│
├── tests/
├── requirements.txt
└── README.md

Installation & Setup Instructions

1. Clone the Repository

git clone https://github.com/dgibbs20/Capstone_Project.git
cd Capstone_Project

2. Create & Activate Virtual Environment

python3 -m venv .venv
source .venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

Running the Application

Step 1: Initialize the Database

cd backend
python db.py

Step 2: Launch Backend API

python -m uvicorn app:app --reload --host 127.0.0.1 --port 8000

Step 3: Open the Frontend

Open:
frontend/index.html
This loads the UI, allowing:

• Transaction entry
• Automatic category prediction
• Saving to database
• End-to-end system interaction

Machine Learning Model

Model: Support Vector Classifier (Linear SVC)
Pipeline: TF-IDF Vectorizer → Linear SVC
Performance: ~90% accuracy
Metrics are stored in:
/backend/metrics/

Capstone Requirements Covered

• Full-stack AI system
• Machine learning model + pipeline
• Frontend UI
• FastAPI backend
• SQLite database
• Documentation
• GitHub repo with full installation instructions
• Ready for instructor testing

Demo & Screenshots

Below are screenshots demonstrating the Smart Budgeting Agent in action.

Application Homepage

Transaction Prediction Example

These images show:

• The working frontend UI
• The prediction interface
• End-to-end system operation

📽️ Project Demo Video

A full live demonstration of the Smart Budgeting Agent is available here:

▶ Demo Video:
https://github.com/dgibbs20/Capstone_Project/releases/download/v1.0.0/Screen.Recording.2025-12-08.at.9.34.02.PM.mov

This video shows:

• Launching the backend server
• Using the frontend UI
• Running predictions
• Saving transactions
• Reviewing activity in real time

Contact

Derrick K. Gibbs-McGlaston
Email: derrickkgibbs@gmail.com
GitHub: https://github.com/dgibbs20