Ensemble Learning Approach for Symptom-Based Diagnosis of Typhoid and Malaria Co-Infection

Abstract

Insufficient number of medical specialists and high costs of medical equipment have vastly contributed to the increase of death rate especially in rural areas of most developing countries. According to Roll Back Malaria there are 300 million acute cases of malaria per year worldwide, leading to more than one million deaths. About 90% of these deaths occur in Africa, majorly in young children. In addition, malaria when tested; a huge number is coinfected with typhoid. Regularly, symptoms of malaria and typhoid fevers do have common physical characteristics and clinicians do have hitches in distinguishing them. In Nigeria the existing diagnostic systems for malaria and typhoid in rural settlements are inefficient thereby making the result to be inaccurate and resulting to treatment of wrong ailments. In addition, some of the rural areas have no access to good medical lab facility for typhoid and malaria diagnosis and in some cases, individuals cannot afford the cost of medical lab facility in their area. Therefore, in this research, machine learning symptom-based typhoid and malaria diagnosis model using XGBoost ensemble learning algorithm is proposed for an improved classification result. Relatively high-performance accuracy was achieved using cross validation on the dataset collected from hospital. Hence the model will be of a great significant use in terms of cost, accurate diagnosis and quality health care services especially in rural settlement as an alternative and a reliable diagnosis of ailments.