📊 Poll Aggregator Engine

A Python-powered data pipeline that ingests raw U.S. polling data, computes rolling weighted averages, and renders interactive visualizations for election and generic ballot tracking.

---

🗂️ Repository Structure

poll-aggregator-engine/
├── update_csv.py          # Data pipeline — rolling average computation
├── ballot_final.csv       # Processed output data with computed averages
└── Visualize.html         # Interactive chart + table frontend

---

⚙️ How It Works

1. Raw Data Ingestion

The pipeline reads a raw CSV of polling data with the following schema:

Column	Description
Pollster	Name of the polling organization
Date	Poll end date
ApprovalRating	Republican / candidate A result (%)
DisapprovalRating	Democrat / candidate B result (%)
Spread	Raw margin between candidates
Grade	Pollster reliability grade

2. Rolling Average Computation (update_csv.py)

The core of the pipeline applies a rolling window average across all polls:

• Reverses the dataframe so averaging flows oldest → newest
• Applies a configurable sliding window (default: 10 polls)
• Computes Avg_ApprovalRating and Avg_DisapprovalRating per row
• Re-inserts the computed columns into the correct position in the schema
• Outputs a clean ballot_final.csv ready for the frontend

window_size = 10  # Configurable: 10 = congress/generic | 6 = state races | 4 = others

for i in range(len(df_rev)):
    window = slice(max(0, i - window_size + 1), i + 1)
    avg_approval.append(round(df_rev['ApprovalRating'][window].mean(), 1))
    avg_disapproval.append(round(df_rev['DisapprovalRating'][window].mean(), 1))

Why reverse then re-reverse? Pandas rolling averages need data in chronological order to correctly weight recency. The dataframe is reversed before averaging, then flipped back so the newest polls appear at the top of the output — matching how the frontend consumes the data.

3. Output Schema (ballot_final.csv)

After processing, the CSV has two new computed columns injected between DisapprovalRating and Spread:

Column	Description
Avg_ApprovalRating	10-poll rolling average — Republican side
Avg_DisapprovalRating	10-poll rolling average — Democrat side

4. Interactive Visualization (Visualize.html)

The frontend reads ballot_final.csv directly in the browser via fetch() and renders:

• Line chart — plots both rolling averages over time using Chart.js, with confidence bands (±0.5%) around each line
• Data table — lists every poll with color-coded margin pills (red = R lead, blue = D lead, intensity scales with margin size)
• Responsive layout — separate rendering paths for desktop and mobile

---

🛠️ Tech Stack

Layer	Technology
Data pipeline	Python 3, pandas
Data format	CSV
Visualization	JavaScript, Chart.js
Styling	CSS3 (glassmorphism, CSS variables)
Deployment	Static HTML — no server required

---

🚀 Usage

Run the pipeline

pip install pandas
python update_csv.py

Update the input/output paths at the top of update_csv.py to point to your CSV file before running.

View the chart

Open Visualize.html in a browser. Make sure ballot_final.csv is in the correct relative path (../polls/2026-generic-ballot.csv by default — update csvPath in the script if needed).

Note: Because the page fetches a local CSV via fetch(), you may need to serve it from a local server to avoid CORS issues:

python -m http.server 8000

Then open http://localhost:8000/Visualize.html

---

📈 Skills Demonstrated

• Data wrangling — cleaning, reshaping, and reindexing DataFrames with pandas
• Rolling window statistics — custom sliding window logic with configurable size
• Schema management — dynamic column insertion at precise positions in a DataFrame
• Data visualization — Chart.js line charts with confidence bands, custom label plugins, and responsive breakpoints
• Frontend/backend integration — Python pipeline output consumed directly by a browser-based chart
• Clean code — no hardcoded magic numbers, configurable parameters, reusable structure

---

📋 Configuration

Parameter	Location	Default	Notes
window_size	update_csv.py line 8	10	Change to 6 for state races, 4 for smaller datasets
csvPath	Visualize.html	../polls/2026-generic-ballot.csv	Update to match your file structure
axisMin / axisMax	Visualize.html	30 / 60	Y-axis range for the chart

---