Introduction
Research is an ongoing process, and scholars follow distinct patterns to solve their defined problems and contribute to the existing body of knowledge. Notably, every analyst attempts to identify gaps in literature through previous documents’ review and maps out the methods and procedures to be used to develop a viable solution. This paper analyzes the article by Anastasopoulos and Whitford and shows how the elements of the research process have been employed to aid in the development of study results and demonstrate future applications.
Problem Definition and Literature Review
The problem definition shapes the research by determining the methods, tools, and analytic procedures used. The volume of data has been increasing at unprecedentedly high levels, presenting a problem to scientists and various practitioners. Anastasopoulos and Whitford (2018) argue that there has been an ongoing conflict between theory and practice in research. This problem forms the basis of their research to identify how machine learning (ML) can provide a solution to data analysis in public administration. The authors did not carry out an extensive literature review, and no section is dedicated to it. However, they relied on previous researchers from the journal of public administration review on data trends and machine learning.
Research Questions and Hypotheses
The research questions contain the specific elements that the scientists want to address in their work. In this article, Anastasopoulos and Whitford (2018) enquire into the application of ML in bridging the gap between theory and practice. Their main question is: How can ML be used to develop new methods, data, and inference to improve public administration research? In their attempt to solve the question, the authors hypothesize that ML algorithms are the answer to current and future generation big data challenges. They further show that although data volume changes significantly, the algorithms will remain relevant for all shifts, making them an ideal solution to the public administration big data problem.
Conceptual Model and Connection to Theory
The interrelationships between concepts and their contribution to the desired solution are essential for reviewing the methods, results, and future implications. The authors focused on several concepts such as the role of ML as a solution to an emerging problem, different applications of supervised and unsupervised ML algorithms, and the link between ML and big data (Anastasopoulos & Whitford, 2018). This conceptual framework is crucial for mapping out the historical applications of ML and implications on future pubic administration data trends. It connects theory to practice by showing that ML algorithms establish a trade-off between bias and variance through performance metrics.
Key Variables and Research and Survey Design
Since the goal was to apply ML algorithms in the public administration sector, the authors sought to understand organizational popularity through users’ brand favorability. The main variables under test in this research were reputation and perception assessed through social media, particularly the Twitter platform (Anastasopoulos & Whitford, 2018). The authors designed the study to apply the human coding techniques to teach the ML algorithm to assess agent popularity. They chose 26402 tweets from the assortment of agencies 26,402. Out of this, 200 comments were chosen at random and given to three people (Anastasopoulos & Whitford, 2018). The coders were expected to group the tweets into relevant categories that would help to determine how popular an agency was on social media. The survey was done randomly to include all relevant tweets. This design was vital because including coders with different levels of understanding is essential for eliminating bias.
Data Collection, Analysis, and Interpretation
The researchers examined the data collected and the organizations from which to gather information after devising a classification scheme. This application contains information on 13 executive agencies within the President of the United States’ Cabinet that use the Twitter network. They compiled 2,000 tweets per organization accessible on Twitter since about March 28, 2018 (Anastasopoulos & Whitford, 2018). They employed the Twitter software programming interface (API) to guarantee an equally represented sample across companies and to profile for innate distinctions in the number of agency tweets. They connected using the R language and extracted the textual data from Twitter messages from each agency.
Following the acquisition of the hand-coded dataset, the next goal was to see if an ML technique could be used to legitimately classify tweets in the bigger database. The data was split into two sets, training, and testing, by mixing it randomly and dividing it into two different datasets. 70% of the data was set aside for learning and 30% for evaluation in this application using a 95% confidence level (Anastasopoulos & Whitford, 2018). This is a standard default method in most computerized statistical tools and is commonly advocated when data is limited. The authors found that organizations directly servicing the wider populace or focused constituencies were worried about their moral image and used social media to portray the moral character to the public.
Conclusion
In conclusion, the research process encompasses several elements that facilitate data gathering, analysis, and interpretation. The article analyzed herein was based on the role of ML in solving big data challenges in future public administration research. Essentially, the volume and complexity of data change unpredictably, requiring an advanced algorithm for analysis. The authors developed an ML algorithm to evaluate agency popularity using social media as the main public communication platform.
Reference
Anastasopoulos, L., & Whitford, A. (2018). Machine learning for public administration research, with application to organizational reputation. Journal of Public Administration Research and Theory, 29(3), 491-510.