Predictive analytics framework for accurate estimation of child mortality rates for Internet of Things enabled smart healthcare systems. (May 2020)

• Record Type: Journal Article
• Authors: Islam, Muhammad; Usman, Muhammad; Mahmood, Azhar; Abbasi, Aaqif Afzaal; Song, Oh-Young
• Is Part Of: International journal of distributed sensor networks. Volume 16:Number 5(2020)
• Publication Date: 2020-05
• Subjects: Machine learning -- child mortality -- predictive analytics -- smart healthcare; Sensor networks -- Periodicals; Intelligent agents (Computer software) -- Periodicals; Multisensor data fusion -- Periodicals; 681.2
• DOI: 10.1177/1550147720928897
• Languages: English
• ISSNs: 1550-1329

Abstract

Globally, under-five child mortality is a substantial health problem. In developing countries, reducing child mortality and improving child health are the key priorities in health sectors. Despite the significant reduction in deaths of under-five children globally, developing countries are still struggling to maintain an acceptable mortality rate. Globally, the death rate of under-five children is 41 per 1000 live births. However, the death rate of children in developing nations like Pakistan and Ethiopia per 1000 live births is 74 and 54, respectively. Such nations find it very challenging to decrease the mortality rate. Data analytics on healthcare data plays a pivotal role in identifying the trends and highlighting the key factors behind the children deaths. Similarly, predictive analytics with the help of Internet of Things based frameworks significantly advances the smart healthcare systems to forecast death trends for timely intervention. Moreover, it helps in capturing hidden associations between health-related variables and key death factors among children. In this study, a predictive analytics framework has been developed to predict the death rates with high accuracy and to find the significant determinants that cause high child mortality. Our framework uses an automated method of information gain to rank the information-rich mortality variables for accurate predictions. Ethiopian Demographic Health Survey and Pakistan Demographic Health Survey data sets have been used for the validation of our proposed framework. These real-world data sets have been tested using machine learning classifiers, such as Naïve Bayes, decision tree, rule induction, random forest, and multi-layer perceptron, for the prediction task. It has been revealed through our experimentation that Naïve Bayes classifier predicts the child mortality rate with the highest average accuracy of 96.4% and decision tree helps in identifying key classification rules covering the factors behind children deaths.