Predictive Analytics for Flood Detection in Sierra Leone: A Proposed IoT and Machine Learning System Design

Abstract

This paper presents a proposed system design combining low-cost Internet of Things (IoT) sensors and a Long Short-Term Memory (LSTM) machine learning model for flood early warning in Sierra Leone. The design uses an ultrasonic water-level sensor connected through a NodeMCU ESP32 microcontroller to a cloud dashboard and data log, paired with an LSTM model trained on national rainfall data to forecast flood risk ahead of time. The paper describes the proposed hardware and software architecture, the dataset and preprocessing approach, and the plan for field testing and evaluation. It is a system design and pre-deployment evaluation plan; it does not report field-tested results.