A Deep Learning-Based Stacked Generalization Method to Design Smart Healthcare Solution

Abstract

Healthcare or health insurance agencies have been in a continuous state of change, especially during the present technology expansion. They are beneath colossal pressure to forecast customer health issue and to generate surplus premium holders which will simultaneously reduce the cost. Examining and utilizing the vast data available are critical for healthcare companies in designing various strategies in the future. Many such healthcare organizations have already moved toward data mining and analytics with data warehouse methodologies and business intelligence with statistical analysis. However, further adaptation is required, as they must use different data from new sources in a blend with the prior sources. To consider this adaptation, predictive analysis technique is proposed. Predictive analysis comprises diverse statistical methods from predictive modeling, machine learning, and data mining that analyze present and past realities to make predictions about future. There are several advantages of using descriptive and predictive analytics in healthcare domain for concrete decision making of cost-effective solutions to their customers. This paper expands upon risk mitigation tactics to foresee high-risk patients. This is done by considering clinical data that is an electronic medical record and evaluating risk associated using stacked ensemble machine learning techniques. This technique helps achieve higher predictive accuracy of 90.17% and specificity of 94.90% in identifying high-risk patients from a lesser amount of data. It has been centered on payer analytics and customer analytics in light of novel machine learning algorithm to give full-cycle knowledge toward cost reduction and improvement in nature of care.