Back to projects

DataShipped2025

CryptoSent

A sentiment-driven market predictor connecting Twitter sentiment to stock price movements.

250k+
Tweets analyzed
24
Tickers tracked
Daily
Data pipeline

Timeline

Feb 2025 to May 2025

Role

Data pipeline and analysis

Team

Solo

Status

Shipped

Stack

PythonpandasNumPySciPySQLSnowflakePower BI

Source

Section 01

Overview

Markets move on mood as much as math. CryptoSent tests how far public sentiment can go as a predictive signal: it scores Twitter sentiment with NLP techniques and correlates it against stock price movements.

FIG. 01 · Sentiment vs. price movement dashboard
Screenshot coming soon

Section 02

Data Pipeline

Tweets are collected, cleaned, and scored in Python using pandas, NumPy, and SciPy. The scored data lands in Snowflake, where SQL handles the joining, grouping, and filtering that aligns sentiment windows with market data.

Keeping the heavy lifting in SQL kept the Python layer simple: score, load, and analyze.

  • NLP sentiment scoring over 250k+ tweets
  • Snowflake warehouse with SQL transformations
  • Daily refresh aligning sentiment windows to trading days

Section 03

Analysis & Visualization

Statistical analysis in SciPy tested whether sentiment shifts lead price movements or just echo them. Findings were visualized in Matplotlib and Power BI, from correlation heatmaps down to per-ticker sentiment timelines.

FIG. 02 · Correlation heatmap, sentiment vs. returns
Screenshot coming soon
FIG. 03 · Per-ticker sentiment timeline
Screenshot coming soon

Section 04

Results

The strongest finding: sentiment spikes correlate with short-horizon volatility more reliably than with direction. The project became my template for data work: a clean warehouse, honest statistics, and visuals a non-technical reader can follow.