Textual Machine Learning: An Application to Computational Economics Research
- PDF / 473,742 Bytes
- 17 Pages / 439.37 x 666.142 pts Page_size
- 78 Downloads / 177 Views
(0123456789().,-volV)(0123456789().,-volV)
Textual Machine Learning: An Application to Computational Economics Research Christos Alexakis1 • Michael Dowling1 • Konstantinos Eleftheriou2 Michael Polemis2,3
•
Accepted: 16 November 2020 Ó Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract We demonstrate the benefit to economics of machine learning approaches for textual analysis. Our use case is a machine learning based structuring of research on computational economics based on 1160 articles published in the Computational Economics journal from 1993 to 2019. Our Latent Dirichlet Allocation approach, popular in the computer sciences, use a probabilistic approach to identify shared topics across a body of documents. This combines natural language processing of article content with probabilistic learning of the latent (hidden) topics that link groups of articles. We show that this body of research can be well-described by 18 overall topics and provide a structure for computational economists to adopt this approach in other avenues. Keywords Topic modeling Latent Dirichlet allocation Computational economics
& Konstantinos Eleftheriou [email protected] Christos Alexakis [email protected] Michael Dowling [email protected] Michael Polemis [email protected] 1
Rennes School of Business, 2 rue Robert d’Arbrissel, 35065 Rennes, France
2
Department of Economics, University of Piraeus, 80 Karaoli and Dimitriou Street, 18534 Piraeus, Greece
3
Hellenic Competition Commission, Athens, Greece
123
C. Alexakis et al.
1 Introduction We use a corpus of research articles to demonstrate the power of machine learning applied to textual learning. Our applied technique is a Latent Dirichlet Allocation (LDA) topic modeling approach, popular in the computer sciences but largely not previously applied in economics. Our purpose, as part of the journal Special Issue on Machine Learning in Economics and Finance, is to demonstrate how machine learning has opened up bodies of text knowledge for analysis and data input in a way that traditional techniques in economics have not easily allowed. We take as a dataset all articles published in the Computational Economics journal and show how we can group these articles based on shared topics obtained through a textual machine learning technique. The main reason for choosing Computational Economics journal articles as our dataset of text documents is reader familiarity, but a second important reason is that it is an interesting text dataset to explore because of the development of the area. The significant advances in computing power in the last decades led to the creation of the new discipline of computational economics. Computational economics is a field of economics at the interface of computer science, economics, and management science. It focuses on the integration of information technology into economics and the automation of manual processes. Specifically, computational economics utilize advanced computing to solve empirical and th
Data Loading...
