10 / Weather data & machine learning
ATMOS INTELLIGENCE
Recognizing weather conditions from observations. ATMOS INTELLIGENCE connects data cleaning, outlier detection, feature analysis, and random-forest classification in a public notebook.
GitHub · View repository ↗RoleTeam member and model development
Period10 months
StatusPublic experiment notebook
The Challenge
How can noisy weather observations become a classification task, while keeping observed-condition classification distinct from forecasting future weather?
Approach
- Clean GlobalWeatherRepository records and normalize weather labels and selected locations
- Inspect outliers with Isolation Forest and explore weather and air-quality features
- Build a 100-tree RandomForestClassifier experiment and inspect feature importance
System Structure
- Data: observation tables, condition labels, and cleaning records
- Features: numeric weather variables, air quality, and outlier screening
- Model: label encoding, random train/test split, and random forest
- Analysis: feature importance, visual exploration, and experiment comparison
Outcomes
- Public README and notebook containing experiment code and outputs
- A documented path from data exploration to random-forest modeling
- An inspectable starting point for classification evaluation and temporal validation
Discuss this project
Interested in this project’s methods, implementation, or tradeoffs? I welcome a conversation by email.