CIS-2450-Kalshi-Term-Project

This repo contains the codebase for our CIS 2450 Term Project on Kalshi Political Markets.

Project Overview

This project aims to analyze and predict outcomes in Kalshi Political Markets.

File Structure + Descriptions

• dashboard/
  • dashboard.py : A Dash app that visualizes model performance and feature importance.
• data/
  • raw/
    • kalshi_markets.parquet : Raw market data pulled from the Kalshi API.
    • kalshi_trades.parquet : Raw trade data pulled from the Kalshi API.
  • intermediate/
• models/ : Saved model artifcats for dashboard display.
• notebooks/
  • 1_eda.ipynb : Exploratory Data Analysis for the raw Kalshi data and GDELT join.
  • 2_feature_engineering.ipynb : Feature engineering and analysis of the all features.
  • 3_modeling.ipynb : Evaluating predictive models for Kalshi market outcomes.
• src/
  • kalshi_api.py : Contains helper functions used for authenticating and making requests to the Kalshi API.
  • pull_markets.py : Fetches all resolved Kalshi political markets (within our time scope) and save to parquet.
  • pull_trades.py : Fetches trades for every political market and save to parquet.
  • build_prediction_targets.py : Builds the prediction_targets.parquet file for the Kalshi data.
  • gdelt_loader.py : Downloads GDELT 1.0 events, entity-links them to Kalshi markets via embedding similarity, and produces per-market sentiment features.
• README.md
• Can the Crowd Be Beaten? Slides.pdf

Gitignore

• secrets\ : stores private Kalshi API key
• .env : environment variables for API key
• *.parquet : files too large for git, stored on Drive