Predicting Elderly Patient Behaviour in Rural Healthcare Using Machine Learning Prince Appiah 1 , Thierry Oscar Edoh 2 [0000-0002-7390-3396] and Jules Degila 3[0000-0003-4688-9178] 1 University of Education Winneba-Kumasi, Ghana Princeappiah35@gmail.com 2 RFW-Universität Bonn, Bonn, Germany oscar.edoh@gmail.com 3 Institute of Mathematics & Physical Science, Porto-Novo, Benin Jules.degila@imsp-uac.org Abstract. The digitization of modern health care data in a rural community has produced a vast amount of patient data stored in health care record systems. Together with the rise of computing power this data could produce effective insight through advanced analysis of this data and include it in medical applications for use in daily operations. This is the case in which structured, semi-structured and unstructured dataset from emergency room admissions is used for machine learning, in order to develop models that predict the possibility of an elderly patient returning to an emergency room within 96 hours. Logistic regression was the selected algorithm since it commonly used in the healthcare data set. The results from the model had a recall of 73% and a precision of 78%. This paper discusses the implementation of such a model in daily operations with a new approach to cost benefits. In other instances, the study is a proof of the concept of predictive modeling in a health care context in rural communities. Keywords: Machine Learning, Rural healthcare, unstructured dataset, Logistic regression 1. Introduction The invention of the computer and even long before that, people are trying to predict or interpret different outcomes from data. This also applies to the health care sector notwithstanding this sector is extremely sensitive to errors because of varied reasons [1]. Rural healthcare centers are exponentially increasing the amount of data. Which has become opportune for machine learning [2]. Data collected in a rural hospital containing vitals, lab results, metadata, etc. can be combined into individual records for every patient. Through machine learning techniques we are able to make use of all this data. Using these algorithms we can predict various outcomes and form recommendations to support medical professionals or do predictions regarding the patient’s health. For example, algorithms can help assign medicines and treatments to patients, they can support medical professionals in making diagnoses and present new origins of certain diseases, the possibilities are endless. This knowledge can also be used to act proactively on different issues. It is, for example, possible to set computerized alerts when specific thresholds are exceeded to prevent unwanted consequences. However, despite the volume of data stored, the potential of using the data for strategic and operational decisions through the means of data analysis is, especially in rural health care, rarely acknowledge. It is obvious that, with this aforementioned data of patients, rural healthcare centers are not using data mining techniques to predict the behavior of elderly patients. The purpose of this study is to apply machine learning to rural health care data, in order to predict elderly patient behavior, providing a basis for medical decisions or risk stratification. The study focus on the prominent property of emergency patients. The fact that ten percent of elderly patients sent from emergency room return within 96 hours. This will help elderly patents to be identified before they return, decisions concerning their care could be taken in Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). IREHI-2019: International Conference on rural and elderly health Informatics, Dakar, Sénégal, December 04-06, 2019