Natural Language-based Financial Forecasting
Natural language-based financial forecasting (NLFF) is a new terminology which becoming the intersection between natural language processing (NLP) and financial forecasting (Xing et al. 2018). NLFF covers several aspects, such as semantic computing, natural language understanding, and time series analysis. According to Xing et al. (2018), the terminology of NLFF was firstly introduced by Nassirtoussi et al. (2014). The growing volume of financial reports, press release, news articles, and social media undirectly becomes the point for NLFF development.
Figure 1 shows the number of scientific papers with topics NLFF in Scopus that had been growing exponentially in the period 1998 – 2016. In Indonesia, NLFF scientific papers are:
- Cakra and Trisedya (2015) had conducted the JCI modelling using features extraction of sentiment and linear regression to the 13 registered stocks on Twitter for 14th – 30th April 2015 and
- Saputro (2019) had conducted the JCI and rupiah exchange rate against the dollar prediction on Twitter using the combination of sentiment analysis and the average of TF-IDF calculation for 27th May – 27th July 2018. The data modelling in Saputro’s research used the least absolute shrinkage and selection operator (LASSO) and model averaging.
One of the challenges in NLFF is sentiment analysis, that is about how to extract and obtain any relevant information from textual data and applied to the financial data analysis (Xing et al. 2018). Several approaches used, for instance, knowledge-based technique, statistical methods, and hybrid approach.
You can directly head over to my Github repository!
Luckily, I did a lot of exploration in this topic for my last year research in Department of Statistics, IPB University. Let’s check this out!