Data Science for Finance Interview.
- 1. James LoBuono’s Interview with the CFA Society Los Angeles (June 2019) Since 2017, CFA Society Los Angeles (CFALA) has hosted “Introduction to Data Science for Finance” classes taught by Cognitir to aid CFALA members in acquiring the basics of data science skills. With a constantly changing and digitally- driven world, the demand for data scientists and data science-related skills is growing rapidly. Glassdoor listed the data scientist job as the “Best Job in America” in 2016, 2017, and 2018, and it expects the demand for data scientists to grow by 20% by 2020. James LoBuono, Global Head of Business Development at Cognitir, as well as one of the company’s instructors, agreed to answer questions about the field and Cognitir’s classes: The topic of data science seems to appear everywhere. Could you talk about the growing demand for data scientists in today’s job market? “Data science is literally applicable to every industry. Technological resources at scale, such as the growing deployment of sensors and cloud computing, allowing for massive amounts of data to be aggregated for analysis, particularly at an unprecedented rate over the past decade. Data science in a nutshell is the science of extracting useful information from data and applying this information to employ sound decision making instead of using just gut, intuition or past experience. The best decisions use a combination of data and gut feel. As a meaningful example, in the financial services world, data science is being utilized through the deployment of machine learning models that work to analyze massive data sets that we can categorize as being “alternative” to traditional information that financial analysts have manually looked at in the past such as annual reports, industry reports, and the major financial statements like balance sheets and income statements. Some of these new forms of information include immense datasets ranging from satellite imagery feeds to social media posts that can only be analyzed effectively with machine learning models in combination with powerful processing architectures, all with the purpose of creating predictive models that drive investment decision making. Analyzing alternative datasets is of course just one example of data science application in the financial world, and these types of decision making framework, based on predictive modeling, are being applied more and more on a daily basis to other fields such as medicine, retailing, and even law. As an example, data science is being used for civil labor law cases to recognize daily employee behavior patterns over the course of many years. This is an analysis that would be pretty hard for one person or a group of people to do on a manual basis, without today’s computing infrastructure.” Why can’t computers do all of the work instead? “Humans still need to be involved and skilled in regards to data science application, because preparing and cleaning data is such a significant component of the process. Complex datasets can be very messy, and depending on where they are sourced from, they will not appear in a neat spreadsheet like traditional finance folks may be used to seeing when they download a 5 year price history of a security off of the web. Before data is fed into a machine learning model where the crunching is done by high powered GPUs or application specific processors, humans may need to engage in the arduous and time consuming process of making sure that data is formatted properly and gauge how complete and valid it is. Many of the real world datasets used in fields such as finance are considered to be unstructured forms of data, in that they are not organized in a formalized manner that is ready to be analyzed. Data that are derived from “big data” sources such as Internet of Things (IoT) sensors as opposed to being housed in a company’s traditional relational database management system (RDBMS) will be directly procured in a messy form, and as such, cleaning and formatting data ahead of getting to the end decision making stage, could consume over 80% or more of the entire process alone. Remember, machines can only do so much! Let’s not forget that data is converted into 1s and 0s for machines to learn and build predictive models on. There is certainly a great deal of ‘gray area’ in data science, and the best data science outcomes leverage the best aspects that humans and machines both bring to the table.” What are some of the tools and skills used by data scientists? “Data scientists and business folks that learn some of the basics of the field, have various methods and tools that they can
- 2. employ to make informative business decisions. One of the most common applications in use today is known as machine learning. Machine learning by simple definition is a way for computers to recognize patterns based on their accessing of data. Machine learning comes in different shapes and flavors. One major example of machine learning, and a methodology that we have dedicated an entire one-day class to is known as Machine Learning Classification. This technique is used to categorize whether or not an observation belongs to a certain class in a binary sense (but can be more than a two-class problem too). The most basic example of classification that we see today in our everyday lives is email being classified as ‘spam’ or ‘non-spam’. Folks working in and around data science utilize Python and R primarily to build data science/machine learning models. At Cognitir, all of our programming courses are taught using Python. Compared to other programming languages, Python is easier to learn and it is backed by a large, globally based development community that equips users with many open- source data science and machine learning libraries that are geared towards specific use cases or applications. Then, there is a world of folks who just use pre-packages data visualization software for their analytics needs. Such software includes Tableau, Microsoft Power BI, Looker, etc. We are biased of course, but we believe that learning a fundamental general purposes computer language such as Python affords the greatest flexibility and horsepower. Cognitir regularly tells finance professionals that learning a programming language like Python is not just for computer scientists. Python can help finance professionals automate tasks on the job. We do not expect these professionals to become computer scientists overnight by any means, but we do expect them to be able to interface more regularly and constructively with technology teams, especially in the context of finding an edge in today’s world. Many large, traditional asset management shops are bringing on lead data science personnel that are becoming directly involved in investment decision processes, and it is pretty self-explanatory that getting down at least some basic data science skills at this juncture is pretty critical.” In the course description, you mention “how data science can help solve a variety of business problems and unlock a plethora of strategic opportunities.” Could you elaborate on some of these business problems that would be relevant to our society’s members along with the same for the strategic opportunities? “On a fundamental level, this harkens back to what data science is at the most practical level. Instead of making strategic business decisions based off of our gut feeling and best guesses, we are now making more and more decisions than ever based on inconceivably large sets of historical data that the human mind would not be able to come remotely close to processing or finding patterns in on its own. Many predictions and forecasts that company management teams used to make based on intuition, can now be made based on verifiable data rooted in science and logic.” As an example, how can today’s technologies, and other forms of automation driven by software, optimize reviewing candidates and the overall hiring process? “This is a great example of technology automating seemingly mundane tasks. Nowadays, many companies employ application tracking systems (ATS) to scan resumes for keywords instead of having a human or group of humans read each application individually. As we tell many of our students, having some of the key buzzwords and in-demand technical skills on their resumes, such as Python, SQL, and data science, are a great way to increase their chances of their resume being reviewed by a decision maker!” Our partnership offers two tracks to help you plan your professional development—programming and non-programming. These will help you grasp knowledge about data science and diversify your portfolios. CFA Society Los Angeles will be hosting Cognitir’s Introduction to Python July 20th .