Authors: Seth Goodman*, William & Mary, Daniel Runfola, William & Mary, Ariel BenYishay, William & Mary
Topics: Spatial Analysis & Modeling
Keywords: machine learning, conflict, convolutional neural networks
Session Type: Paper
Start / End Time: 1:45 PM / 3:00 PM
Room: Director's Row I, Sheraton, Plaza Building, Lobby Level
Presentation File: No File Uploaded
Machine learning applications have a growing role in the monitoring and evaluation of international development. Recent work has shown the utility of combining widely available satellite imagery with convolutional neural networks (CNNs) to produce estimates of sparse data such as a poverty rates at relatively granular levels. This paper will explore the expansion of CNN based methods to predict the likelihood of conflict in Nigeria, incorporating multi-spectral Landsat 8 imagery and pre-trained convolutional neural networks. The Armed Conflict Location & Event Data (ACLED) dataset will be used for fine-tuning and validation of the training of the CNNs over a range of hyperparameters and sampling schemes. Indicators of future conflict in developing countries have the potential to be a valuable resource for decision makers sending personnel and resources to conflict prone regions.