alexiskadje-ai/MEng-Dissertation

This is my Masters of Engineering dissertation titled 'PREDICTIVE ANALYTICS FOR CROP YIELD MODELLING CASE OF MAIZE IN THE WESTERN HIGLANDS OF CAMEROON # Maize Yield Prediction using Multi-Source Data Integration Overview This repository contains the complete implementation of my Master's dissertation: "PREDICTIVE ANALYTICS FOR CROP YIELD MODELLING CASE OF MAIZE IN THE WESTERN HIGHLANDS OF CAMEROON". The research develops a hybrid machine learning framework that integrates multi-source geospatial and agricultural data to predict maize yields with uncertainty quantification. Research Objectives - Develop a hybrid ConvLSTM-XGBoost model for spatio-temporal yield prediction - Integrate heterogeneous data sources (satellite, climate, soil, yield records) - Quantify prediction uncertainty and model interpretability - Identify key yield determinants in the Western Highlands agro-ecological zone - Deploy an interpretable framework for agricultural decision support System Architecture The hybrid architecture combines: - ConvLSTM: For spatio-temporal patterns from satellite time-series - XGBoost: For tabular features from climate and soil data - Attention-based Fusion: For optimal feature combination - Uncertainty Quantification: Bayesian methods for prediction confidence Dataset Description | Data Source | Parameters | Resolution | Period | |-------------|------------|------------|---------| | Satellite Imagery | NDVI, EVI, LST indices | 10-30m | 2020-2024 | | Climate Data | Rainfall, Temperature, Humidity, Solar Radiation | Daily | 2018-2024 | | Soil Properties | pH, N, P, K, Organic Matter | 120 sites | 2020-2024 | | Yield Records | Maize production statistics | Regional | 2018-2024 | Data Sources - Satellite: Sentinel-2 (10m), Landsat 8 (30m) - Climate: NASA POWER, Ground Weather Stations - Soil: Laboratory measurements from 120 sampling sites - Yield: MINADER official statistics + 45 farm ground-truth records Installation & Setup Prerequisites - Python 3.8+ - Jupyter Notebook - LaTeX (for dissertation compilation) Installation 1. Clone the repository bash git clone github.com …