A Cluster-Based Machine Learning Ensemble Approach for Geospatial Data: Estimation of Health Insurance Status in Missouri

Abstract

Mainstream machine learning approaches to predictive analytics consistently prove their ability to perform well using a variety of datasets, although the task of identifying an optimally-performing machine learning approach for any given dataset becomes much less intuitive. Methods such as ensemble and transformation modeling have been developed to improve upon individual base learners and datasets with large degrees of variance. Despite the increased generalizability and flexibility of ensemble approaches, the cost often involves sacrificing inference for predictive ability. This paper introduces an alternative approach to ensemble modeling, combining the predictive ability of an ensemble framework with localized model construction through the incorporation of cluster analysis as a pre-processing technique. The workflow not only outperforms independent base learners and comparative ensemble methods, but also preserves local inferential capability by manipulating cluster parameters and maintaining interpretable relative importance values and non-transformed coefficients for the overall consideration of variable importance. This paper demonstrates the ensemble technique on a dataset to estimate rates of health insurance coverage across the state of Missouri, where the cluster pre-processing assists in understanding both local and global variable importance and interactions when predicting high concentration areas of low health insurance coverage based on demographic, socioeconomic, and geospatial variables.

Extending cluster-based ensemble learning through synthetic population generation for modeling disparities in health insurance coverage across Missouri

“…The algorithms developed, along with methods for determining which variable, or combination of variables, to use as the clustering variable, were demonstrated on a dataset where the objective was to determine the algorithms' ability to predict the distribution of health insurance status across the state of Missouri using demographic, socioeconomic, access, and relative-location information as independent variable loadings. Results from the research found that using a cluster-based ensemble approach outperformed all other modeling techniques evaluated in the study, including linear and nonlinear base-learning algorithms and three aggregate ensemble techniques [15]. In addition to improved predictive performance, the cluster-based ensemble also provided increased inferential power in assessing relative variable importance.…”

“…Building on methods from Trivedi et al [23] and previous research of Mueller et al [15] that examined the relationship between cluster analysis as a preprocessing technique and predictive performance associated with various base learner and ensemble machine learning models, this study sought to enhance algorithmic performance of the ensemble methodology for geospatial data through a technique known as synthetic population generation. The algorithms developed, along with methods for determining which variable, or combination of variables, to use as the clustering variable, were demonstrated on a dataset where the objective was to determine the algorithms' ability to predict the distribution of health insurance status across the state of Missouri using demographic, socioeconomic, access, and relative-location information as independent variable loadings.…”

“…Research objectives outlined for this study were direct extensions of previous research, where stability and sensitivity testing sought to validate the previously developed cluster-based ensemble technique while expanding the method by providing an improved workflow for examining global as well as localized variable importance. Mueller et al [15] demonstrated the newly developed clusterbased ensemble method on a single dataset, examining health insurance coverage with carefully chosen independent variables backed by qualitative theory relating demographic and socioeconomic information, physical location, and access to care, to the distribution of individuals in Missouri lacking health insurance. While the cluster-based ensemble outperformed other ensemble and base-learning algorithms for the health insurance dataset, the first objective in this study seeks to better determine whether the algorithm can be applied to other datasets where the underlying characteristics may vary in comparison to the dataset used in the original study.…”

The Changing Narrative in the Health Insurance Industry: Wearables Technology in Health Insurance Products and Services for the COVID-19 World

“…It would be tautological to argue that such data would be utilised subsequently to determine the subsequent set of premiums (once a customer was onboard) to be paid by the customers. This could be carried out based upon machine learning (ML) or artificial intelligence (AI) based predictive analytics (Kose et al, 2015; Kuo et al, 2007; Mueller et al, 2019). Once the factor inputs data (a large set of it) was available, ML and AI could help HI managers to ascertain the most appropriate premium values.…”

Tri-Clustering Based Exploration of Temporal Resolution Impacts on Spatio-Temporal Clusters in Geo-Referenced Time Series

“…As an important data mining task, clustering is useful for exploring patterns in GTS by assigning similar data elements into the same cluster and dissimilar elements into different ones [7,8]. As a result, it provides both an overview of data at cluster levels and investigation of details on single clusters [9,10].…”