💸 DoomSpend — AI-Powered Expense Intelligence Dashboard

Turning messy transactions into clear financial decisions.

---

🚀 Live Demo

👉 https://doomspend.streamlit.app/

---

🎯 Why This Project Exists

Tracking expenses sounds simple — until real life happens.

Transactions like:

• “zomato order”
• “uber ride”
• “gpay to friend”

…are messy, inconsistent, and hard to analyze.

➡️ Result:

• No clarity
• No patterns
• No control over money

DoomSpend solves this by converting raw transaction text into meaningful financial insights.

---

🧠 What It Does

DoomSpend is an end-to-end ML-powered financial intelligence system that:

✔ Understands raw transaction text ✔ Automatically classifies expenses ✔ Visualizes spending behavior ✔ Detects anomalies ✔ Suggests savings opportunities

---

✨ Features

Feature	Description
🧠 NLP Classification	Predicts category from text input
📊 Interactive Dashboard	Real-time financial insights
📈 Trend Analysis	7-day moving average visualization
⚠️ Anomaly Detection	Flags unusual transactions
💡 AI Recommendations	Suggests savings improvements (~₹98K)
💰 Cash Flow Analysis	Income vs Spending visualization

---

🧩 System Architecture

User Input
   ↓
Text Cleaning & Normalization
   ↓
TF-IDF Vectorization (Bigrams)
   ↓
Naive Bayes Model
   ↓
Prediction
   ↓
Streamlit Dashboard
   ↓
Insights + Anomaly Detection + Recommendations

---

📸 Screenshots

📊 Dashboard Overview

📈 Spending Distribution

💰 Cash Flow Health

---

📊 Dataset

• 1000+ transactions

• Simulates real student financial behavior

• Categories:

  • Food 🍔
  • Travel 🚗
  • Bills 💡
  • Shopping 🛍️
  • Entertainment 🎬
  • Income 💰

• 📅 Time Range: Jan – May 2026

• Includes noise, ambiguity, and real-world inconsistencies

---

🧪 Model & Approach

🔹 Feature Engineering

• TF-IDF with bigrams

• Captures contextual meaning:

  • “uber ride” ≠ “uber eats”

---

🔹 Model

Multinomial Naive Bayes

• Optimized for text classification
• Efficient on sparse data
• Fast and scalable

👉 Also evaluated Logistic Regression as a baseline

---

📈 Results

• 🎯 Accuracy: ~85%
• 📊 Balanced performance across categories
• 🧠 Handles noisy real-world inputs effectively

---

🧠 Key Insights

• Travel & entertainment → highest spending drivers
• Food → frequent but low-value transactions
• Income → sparse but high-value

💡 Potential savings improvement: ~₹98K

---

⚠️ Anomaly Detection

• Flags transactions > 1.8× category average
• Based on rolling 30-day behavior

---

💻 Tech Stack

• Python
• Pandas, NumPy
• Scikit-learn
• Streamlit

---

⚡ Run Locally

git clone https://github.com/your-username/DoomSpend.git
cd DoomSpend
pip install -r requirements.txt
streamlit run app.py

---

🚧 Limitations

• Synthetic dataset
• Limited vocabulary scope
• No real banking integration

---

🔮 Future Improvements

• BERT / Deep Learning models
• Real-time transaction ingestion
• Mobile/Web deployment
• Personalized financial AI

---

🏁 Final Thought

DoomSpend goes beyond expense tracking — it transforms financial data into decision-making intelligence.

---

👩‍💻 Author

Riya Shah B.Tech Computer Engineering

---

⭐ If you like this project

Give it a ⭐ on GitHub — it helps a lot!