Machine Learning Algorithms for Big Data Applications With Policy Implementation

Introduction

Policy implementation is a critical part of public policy, referred to as the process of carrying out an underlying policy decision, typically made in a statute (Sabatier & Mazmanian, 1980). Researchers have always paid much more attention to policy design and policy evaluation and less on how to implement these policies (Schofield, 2001). The first studies of implementation theory occurred in Oakland, California (Pressman & Wildavsky, 1984). Since then, there has been a rapid rise in the study of policy implementation. Previous work has focused on critical analysis or synthesis of different approaches (Sabatier, 1986). However, policy implementation research remains a niche area of interest.

Given that many studies of policy implementation provide evidence of the importance of implementation, the aim of this study was to further current knowledge of their rigidities. Based on selected published research, this research combines the application of the text mining technique with a structural topic model (STM) to provide a snapshot of policy implementation studies from January 2000 to July 2019 as well as incorporating information about the documents (Roberts et al., 2013). Within the literature analysis, we have tried to highlight the manifestation of the rigidities in policy implementation study development to facilitate future research (Moro et al., 2019). Furthermore, the STM could allow the researcher to discover topics from textual data without predicting them.

Text mining and topic modeling have been successfully used as a valuable tools in research on ethnic marketing (Moro et al., 2019) and cause-related marketing (Guerreiro et al., 2016). The generally accepted use of advanced text mining methodology could provide a comprehensive data-driven analysis of research (Liu et al., 2018). We chose unsupervised algorithms because we had no a priori expectations for categories of scholarly study, especially on topics some researchers might not have considered (Reich et al., 2014). The unsupervised learning model could help researchers analyze textual data without human intervention. Until now, this methodology has not been applied to the policy implementation literature. In this paper, we demonstrate policy implementation research using topic models instead of focusing on the technical details of the STM.

The main contributions of this study are as follows. First, we uncovered the prevalence and content of topics regarding policy implementation, based on text mining and the STM. Second, we explored the changing trends regarding the proportion of scholarly attention to different topics. Third, we evaluated the levels of scholarly impact by each topic. The rest of this paper is broken into four sections. The next section gives a brief overview of policy implementation, after which we outline text mining and an STM that includes covariates. We then present our results before, finally, drawing our conclusions for policy implementation research.