Identifying the influential factors of commodity futures prices through a new text mining approach

Abstract

How to determine the numerous and complicated factors that affect the dynamics of commodity futures prices is still a great challenge. The existing studies have mainly identified influential factors based on researchers’ judgements or the summarization of previous studies, which is relatively subjective and makes it difficult to attain comprehensive factors. This paper proposes a new text mining method named Dependency Parsing-Sentence-Latent Dirichlet Allocation (DP-Sent-LDA) to identify the influential factors of commodity futures prices objectively and comprehensively from a massive number of news headlines. In the empirical analysis, based on 49 501 news headlines about six Chinese commodity futures over the period of 2011–2018, a total of 104 specific influential factors are identified, and their relative importance is given. The identified influential factors not only contain almost all the widely studied influential factors but also include some factors rarely mentioned in the existing literature. Regression analysis is conducted to validate the effectiveness of these influential factors.

Keywords:

• Text analysis
• Dependency Parsing-Sentence-Latent Dirichlet Allocation (DP-Sent-LDA)
• Topic modelling
• Commodity futures prices
• Financial risk

JEL Classification:

• E44
• G12
• G13

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This study was supported by grants from the National Natural Science Foundation of China (71425002, 71971207, and 71601178) and the Youth Innovation Promotion Association of Chinese Academy of Sciences (2012137 and 2017200).