victorkelechi/Indigenous-Weather-ML

A machine learning project predicting rainfall (heavy, moderate, small) in Ghana’s Pra River Basin using Indigenous Ecological Indicators (sun, clouds, wind, moon, plants, animals). It digitizes traditional knowledge via SIW App, merging local wisdom with AI for inclusive, hyper-local weather forecasting. # Indigenous-Weather-ML A machine learning project predicting rainfall (heavy, moderate, small) in Ghana’s Pra River Basin using Indigenous Ecological Indicators (sun, clouds, wind, moon, plants, animals). It digitizes traditional knowledge via SIW App, merging local wisdom with AI for inclusive, hyper-local weather forecasting. ### Overview This project develops a machine learning model to predict rainfall intensity (heavy, moderate, or small) in Ghana’s Pra River Basin using Indigenous Ecological Indicators (IEIs) such as sun, clouds, wind, moon, plants, and animal behavior. By digitizing forecasts collected via the SIW Mobile App and combining them with measured rainfall data, this work explores how AI can validate and strengthen indigenous climate wisdom. The goal is to build inclusive, hyper-local weather prediction models that empower rural farmers, improve agricultural planning, and bridge cultural and scientific knowledge systems. ### Data Collection The dataset was collected using the Smart Indigenous Weather App, a custom-built mobile app available on the Google Play Store. Key steps in data collection included: > Rain Gauge Deployment: Calibrated rain gauges were installed on the properties of 25 local farmers across three regions (Central, Eastern, Ashanti). > Farmer Training: Farmers were trained to record daily rainfall observations and to submit indigenous forecasts using the app. > Use of Indigenous Ecological Indicators (IEIs): Farmers made forecasts based on traditional signs such as cloud formations, sun position, wind, moon, heat, and specific tree or animal behaviors. Variables collected include: > Daily rainfall amounts (via rain gauges). > Forecast type (heavy, moderate, small, or no rainfall). > Ecological indicators used (cloud, sun, moon, heat, wind, plants, animals). > Prediction time frame (12-hour and 24-hour forecasts). > Accuracy of predictions (hit/miss rate vs actual rainfall). ### Objective The objective of this challenge …