Urban slums in Somalia face significant environmental challenges due to poor waste management, water scarcity, and air pollution. These issues not only affect public health but also hinder economic development. A hybrid machine learning approach combining deep neural networks with reinforcement learning was employed to optimise the IoT sensor network configuration. Data from existing sensors and simulations were used to validate system performance under varying conditions. The optimised sensor array demonstrated a $R^2 = 0.95$ in predicting waste levels, indicating high accuracy in real-world deployment scenarios where data variability was significant. The developed IoT solutions showed potential for enhancing environmental monitoring and management in urban slums of Somalia with lower operational costs compared to traditional methods. Further research should focus on integrating community feedback into the design process and exploring energy-efficient sensor technologies for widespread deployment. Internet of Things, Urban Slums, Environmental Monitoring, Machine Learning, Low-Cost Solutions