Personal CPI Inflation Tracker

Built with Python, SQL, BLS API, and Streamlit

This project computes a Personalized Consumer Price Index (CPI) for each user based on their unique spending behavior. Instead of relying on the national “CPI-U,” this system builds a user-specific inflation index using:

• Cleaned transaction data
• Fixed-weight category baskets (Laspeyres method)
• Normalized CPI data from the BLS
• SARIMAX forecasting models
• An interactive Streamlit dashboard

This behaves like a miniature quant research pipeline: data engineering → economic modeling → time-series forecasting → visualization.

Features

1. Transaction Ingestion & Cleaning

• Load multi-user transaction CSVs
• Clean timestamps, categories, and amounts
• Store clean data in transactions_raw

---

2. SQL-Based CPI Weight Modeling

• Map raw categories → CPI-like buckets
• Compute monthly spending weights
• Identify each user’s base month
• Build fixed Laspeyres weights, same method used for official CPI

---

3. Official CPI Fetch & Normalization

• Pull CPI time-series from the BLS API
• Normalize each category’s index to base = 100
• Store results in cpi_series and cpi_norm

---

4. Personalized CPI Computation

• Combine user weights with normalized CPI
• Generate multi-user time series stored in personal_index

---

5. SARIMAX Forecasting

• Fit SARIMAX(1,1,1)x(0,1,1,12) per user
• Fallback to naive forecasts for short histories
• Save 12-month forecasts with confidence bands

---

6. Interactive Streamlit Dashboard

Includes:

• Personal CPI vs official CPI-U
• Category weight evolution
• CPI forecast
• “What-if” scenario tool (e.g., gas +20%)
• Random User Selector
• Preview of recent transactions

---

Example Visualizations

1. Personal CPI vs Official CPI-U

Shows whether your inflation moves differently than national CPI-U

2. Category Spending Weights Over Time

Highlights how grocery, gas, dining, shopping, and other categories shift over time.

3. Personal CPI Forecast with Confidence Intervals

Predicts personal inflation for the next 6–12 months with confidence intervals.

---

Why This Project Matters

Official CPI uses fixed category weights (Shelter ~34%, Energy ~7%, Food ~13%). But our own spending is different so our inflation is different.

This project captures that by computing:

• Your personal inflation exposure
• The categories driving your cost of living
• Forward-looking CPI projections

It mirrors how analysts build:

• Custom inflation baskets
• Household CPI indexes
• Real return calculations
• Regional inflation models

This project focuses on inflation measurement and scenario analysis rather than trading or alpha generation.

Technology Stack

• Python: pandas, SQLAlchemy, statsmodels (SARIMAX), Plotly, Streamlit
• SQL: SQLite, window functions, category mapping, weight calculations
• APIs: BLS CPI API
• Visualization: Matplotlib & Streamlit interactive dashboards

---

Data Source

This project uses a public, multi-user credit card transaction dataset from Kaggle to simulate household spending behavior. The dataset includes transaction timestamps, amounts, and merchant metadata suitable for CPI-style aggregation.

---

Project Architecture

.
├── data/raw/                     # Raw transaction CSVs
├── src/
│   ├── etl_transactions.py       # Load & clean raw data
│   ├── categorize_sql.sql        # Map to CPI-like categories
│   ├── index_sql.sql             # Compute category weights + base weights
│   ├── Personal_cpi_sql.sql      # Normalize CPI + compute personal CPI
│   ├── bls_api.py                # Fetch official CPI from BLS API
│   ├── forecast.py               # SARIMAX forecasts per user
│   ├── make_charts.py            # Static visualization
│   └── app.py                    # Streamlit dashboard
├── charts/                       # PNG visualizations
├── DB/personal_cpi.db            # SQLite database
├── .env                          # DB_URL and (optional) BLS_API_KEY
├── requirements.txt
└── README.md

---

Setup Instructions

1. Clone the repository

git clone https://github.com/YOUR-USERNAME/Personal-CPI-Tracker.git
cd Personal-CPI-Tracker

2. Create a virtual environment

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

3. Install dependencies

pip install -r requirements.txt

4. Create your .env file

Inside the root directory:

DB_URL=sqlite:///personal_cpi.db
BLS_API_KEY=your_key_here 

5. Load raw transactions

python src/etl_transactions.py

6. Run SQL preprocessing

Open your DB and run:

    categorize.sql

    weights.sql

    cpi_math.sql

7. Pull CPI data

python src/bls_api.py

8. Generate Forecasts

python src/forecast.py

9. Run the dashboard

streamlit run src/app.py