23 – M4 L1B 24 NLP Used To Enhance Fundamental Analysis V1

Advanced natural language processing techniques are being used to enhance fundamental analysis. Fundamental analysis is the kind of work done by traditional investment analysts who work for financial institutions, like mutual funds and brokerage firms. Much of the text-based analysis that fundamental researchers perform on companies, whether it’s reading financial news, reading quarterly earnings reports, listening to earnings calls, or reading forms submitted to regulatory agencies, this work can be enhanced with natural language processing. For example, a human analyst usually tracks 10 to 20 companies. Computers could act as a first-pass filter sifting through the same data source for thousands of companies. The automated methods could filter, categorize, and label the information. This could help fundamental analysts decide which companies or information sources to prioritize for further analysis. So, the pre-processing performed with NLP could support and enhance the existing workflow of fundamental research analysts. One can also look for insights in the required paperwork that companies send to government regulators. In the United States, the Securities Exchange Commission or SEC receives 10-K forms from companies once a year and three 10-Q forms per year. The 10-K details the companies view of its business, its past financial results for the year, as well as its current or potential future business risks. One potential use of these 10-Ks is to track how a company’s business evolves over time. For instance, one study used natural language processing to analyze Amazon’s 10-Ks from the years 2007 to 2016. The researchers attempted to determine which sectors the company most closely resembled based on its 10-K reports. In 2007, the company’s description of its business most resembled the software and hardware sectors. Over time, the company’s description based on it’s 10-K forms resembled more of the retail and software sector and to some extent the media sector. You may be wondering how we could make use of this information. The sector that companies closely resemble are factors that help explain the price movement of their stocks. You could imagine that retail stocks may be correlated to some extent, and technology companies may also exhibit similar price movements. In a later lesson, we’ll learn how the sector is a particularly important factor and we’ll learn about adjusting a portfolio to be sector neutral. Another application of natural language processing on 10-K forms is sentiment analysis. You may be familiar with sentiment analysis on movie reviews or restaurant reviews, usually to categorize user-generated text as positive, neutral, or negative. Sentiment analysis can be used with other categories as well. For instance, one could apply sentiment analysis to estimate how much little risk a company’s 10-K describes or how much uncertainty the company faces from competitors, customers, or suppliers. One could imagine creating factors based on positive outlook, negative outlook, little risk, business uncertainty, or any number of categories. These labels could also be used to create Alpha factors. In fact, you will be applying natural language processing on corporate documents in term two of this program. For now, we want to first build-up a comfortable foundation in factor risk modeling and Alpha factor research. This will give you the context to understand where NLP and deep learning techniques can fit into the quad workflow.

%d 블로거가 이것을 좋아합니다: