1st Place Winner Quadreal AI Challenge 🏆 Using ML to predict mission IoT sensor data
Devpost Link: https://devpost.com/software/quadreal-challenge
- This is our submission to Quadreal AI Challenge organized by uWaterloo Data Science Club
- In this Kaggle-style competition the goal is to ML to predict mission IoT sensor data
- Using a combination of clever feature engineering techniques and XGBoost we won the first place prize in this competition
- We devise a solution to predict missing values for IAQ sensor data in the occasion of outages.
- We also propose a solution for detecting anomalies to be used by sensors to flag abnormal air conditions.
- Doing Analysis on Time Series Data: Analyzing time series data about trends, seasonality, cyclical etc.
- K-Fold Mean Target Encoding: This feature made the biggest different in our MSE score.
- XGBoost: Our versatile classification model.
- Isolation Forest: Useful for anomaly detection.
From the competition server:
