Article Preview
Top1. Introduction
Sickle cell disease (SCD) is a genetic blood disorder affecting red blood cells, with high morbidity and mortality rates. The United Nations has recognized SCD as a global public health concern, and the World Health Organization (WHO) recommends that 50% of member states will have established SCD control programs by 2020 (World Health Organization, 2006). The United Nations General Assembly recognized SCD as a global public health concern due to the morbidity and mortality caused by the disease and the significant social and economic impact that results (United Nations General Assembly, 2009). SCD includes a series of pathological genotypes resulting from the inheritance of Sickle haemoglobin (HbS), a structural variant of normal adult haemoglobin (HbA) (Chakravorty & Williams, 2015).
SCD affects 20–25 million people globally, and it is estimated that 75–85% of children born with SCD are born in Africa, where mortality rates for those under age 5 range from 50% to 80% (Aygun & Odame, 2012; Makani et al., 2011). In sub-Saharan Africa, it is estimated that 240,000 children are born with SCD annually (Makani et al., 2011). Of the 330,000 babies born with a major hemoglobinopathy worldwide, 275,000 have SCD, making it the major global hemoglobinopathy (Aygun & Odame, 2012; Modell & Darlison, 2008; Weatherall, 2011). SCD patients in the developed world account for only 10% of the world’s SCD patient population (Aygun & Odame, 2012). Aliyu et al. (2008) reported that there are between 20 and 25 million people worldwide living with SCD, of which 12–15 million live in Africa. The highest prevalence of sickle-cell trait (SCT) in Africa occurs between the latitudes of 15¡ North and 20¡ South, where the prevalence ranges between 10% and 40% of the population (Agasa et al., 2010).
Machine learning (ML) is a branch of artificial intelligence that employs a variety of statistical, probabilistic and optimization tools to learn from past examples and afterwards use the prior training to classify new data, identify new patterns or predict novel trends (Mitchell, 1997). Machine Learning is used for medical research to generate knowledge from complex data which is helpful for improving clinical decision making process. Presently, machine learning provides several techniques that have been successfully employed to construct predictive models (Jaree et al., 2013). Machine learning can be broadly classified into Supervised and Unsupervised ML. Supervised learning is a technique used for developing prediction models from training data unlike the unsupervised where the model is fit to observation (McClendon et al., 2015). Supervised learning techniques can be used in developing both classification and regression models.
Classification model is a predictive modeling approach aimed at allocating a set of input record to a discrete target class unlike regression which allocates a set of records to a real value (Balogun, 2016). Machine learning has been applied in the area of healthcare to provide information about disease risk, prognosis and survival using predictive modeling. Predictive modeling in healthcare has shown success in the prediction and diagnosis of various critical diseases as well as decision making processes to increase the accuracy of diagnoses and give answers to physicians about individual patients (Jaree & Vatinee, 2015).