AnalyticEngine190/africa-air-quality-forecasting-Nakuru

Forecasting air quality in African cities to identify and analyze public health trends. Forecasting Air Quality in Nakuru (An ARIMA Time Series Story) Hello! Welcome to my project. This was a complete, top-to-bottom data science challenge where I rebuilt one of my favorite projects from my CV: forecasting air quality in Africa. I started with a raw, messy .csv file and ended with a fully tuned, high-performance ARIMA model. This notebook is the story of that journey, showing how I battled messy data, found hidden patterns, and strategically built a model that works. Data Source The data for this project was downloaded from open.africa (the sensors.AFRICA archive), a fantastic open-source platform for air quality data. Direct Link to Data: sensors.AFRICA Air Quality Archive - Nakuru (You can download it from Open Africa November 2025 Sensor Data Archive) Raw File Used: Nakuru.csv (downloaded from the source above) The Final Result I'm thrilled with how this turned out. After all the cleaning and tuning, I built a final model that improved predictive performance by 26.2% over a simple baseline. Model MAE (Test Set) Improvement Baseline (Lag=1) 10.74 - Final Tuned ARIMA(24, 0, 2) 7.92 26.2% This project shows my ability to take a real-world problem and see it through, from identifying key patterns (like the 24-hour cycle) to building a robust model that delivers real, measurable improvements. My Project Journey (The "How-To") The complete, step-by-step analysis (with all my "Aha!" moments) is in the air-quality-analysis.ipynb notebook. Here’s the short story: Battling the Data: The raw .csv file was a mess! The first challenge was just loading it. I had to battle ParserErrors by figuring out the file used semicolons (;) as separators and commas (,) for decimals. I also found corrupt text (like "17.?5") that I had to handle. Preparing for Time Series: Once clean, I resampled the high-frequency data into stable 1-hour averages, converted the timezone to 'Africa/Nairobi', and made sure all my data types were correct. The "A-ha!" Moment: The big breakth …