Fintech in Investment Management: Big Data and AI | CFA Level I Portfolio Management

In this lesson, we’ll dive into Big Data, Data Science, and the advances in AI and machine learning. You don’t need to be a data scientist or understand the technical aspects, but rather grasp the key concepts, their applications in finance, and the challenges of adopting these technologies. So, let’s roll right in!

What is Fintech?

Fintech refers to technology-driven innovation applied to the financial services industry. Companies developing such technologies are known as fintech companies. Some primary areas where fintech is evolving include:

• Handling extremely large datasets for investment decision-making
• Artificial intelligence for analyzing large data sets
• Automated trading through computer algorithms
• Automated advice from robo-advisors for retail investors
• Financial record-keeping using distributed ledger technologies like blockchain

Now, let’s start with Big Data!

Big Data: Characteristics and Sources

Big Data refers to all potentially useful information generated in the economy, including traditional sources (financial markets, company financial reports, economic data) and alternative data from non-traditional sources like:

• Individuals generating data (social media posts, online reviews, emails, web searches)
• Business processes (credit card transactions, direct sales information, corporate exhaust)
• Sensors (smartphones, cameras, RFID chips) part of the Internet of Things

The three main characteristics of Big Data are volume, velocity, and variety. To extract valuable information from Big Data, data science techniques are employed.

Data Science: Processing and Visualization

Data science involves processing and visualizing data through methods like capture, curation, storage, search, and transfer. Visualization techniques vary based on the structure of data, with charts and graphs for structured data and word clouds or mind maps for unstructured data.

Data science presents challenges like ensuring data quality, dealing with unstructured data, and depending on the skill of the data scientist.

Artificial Intelligence and Machine Learning

Artificial intelligence (AI) aims to simulate human cognition using computer systems. Neural networks, for instance, process information in a way similar to the human brain. Over time, neural networks have evolved into machine learning (ML), where algorithms improve their cognitive skills as they’re used.

Machine learning involves two primary types of learning:

Supervised Learning

In supervised learning, both input and output data are labeled, allowing the algorithm to learn how to map inputs to their desired outputs. After the training phase, the ML algorithm is given new data to predict outcomes or recognize patterns. For example, a stock prediction machine can be trained using past data like price history, company fundamentals, and economic data as inputs, and past returns of various stocks as output training data. The trained model can then predict future stock returns using the latest data.

Unsupervised Learning

In unsupervised learning, input data is not labeled. Instead, the machine learns to describe the structure of the data. For example, characteristics of various companies can be input, and the algorithm groups the companies into peer groups based on certain similarities other than standard sector or country groupings.

However, machine learning faces challenges like overfitting, underfitting, and the “black box” problem:

Overfitting

Overfitting occurs when the machine learns the input and output data too precisely, treating noise as true parameters and identifying false or unsubstantiated patterns and relationships. This leads to a model that performs well on training data but poorly on new, unseen data.

Underfitting

Underfitting happens when the machine fails to identify actual patterns and relationships, treating true parameters as noise. In this case, the model is not complex enough to describe the data accurately, resulting in poor performance both on training and new data.

The “Black Box” Problem

A further challenge with machine learning is its “black box” nature, producing outcomes based on relationships that are not readily explainable. This can make it difficult to understand the rationale behind the algorithm’s decisions and predictions.

Now that we have delved deeper into AI and machine learning, we can move on to the next topic, exploring more applications of fintech in investment management. See you in the next lesson!

Deep Learning: A Growing Trend

Deep learning is a popular form of machine learning that uses layers of neural networks to identify patterns, starting with simple patterns and advancing to more complex ones. Applications of deep learning include image and speech recognition.