“How Computers Are Transforming Wall Street Trading and Decision-Making”

### The Importance of Computer Science in Finance: Sentiment Analysis and Machine Learning

#### Introduction
Throughout the years, the domain of computer science has revolutionized various industries, including finance and economics. Presently, leading financial institutions such as JP Morgan, Barclays, and others depend significantly on computational technologies to assess risks, forecast asset movements, and guide investment strategies. As traditional trading floors decline, the prominence of quantitative analysts (“quants”) and machine-learning models has increased, with computation quietly driving a substantial part of Wall Street’s achievements.

So, in what ways do machines and algorithms assist these companies? This article investigates two key methodologies: **sentiment analysis** and **machine learning**, to examine how computers facilitate financial market assessment and improve investment decision-making.

—

### Sentiment Analysis: Grasping Market Sentiment

In the past, stock market analysis often relied on gut feelings. However, contemporary investing focuses on harnessing information. The shift from instinctual decision-making to algorithmic trading has led to the emergence of specialized positions such as quants, whose role is to develop mathematical and machine-based algorithms to enhance trade execution. Central to such an algorithm is “sentiment analysis.”

#### What is Sentiment Analysis?
Sentiment analysis entails assessing public sentiment—including news, remarks, and social media—regarding a specific stock, industry, or corporation. For example, by evaluating articles and financial documents, computers interpret whether the sentiment associated with a stock is **bearish** (negative) or **bullish** (positive). This empowers traders to make swift and well-informed choices.

#### The Role of Lexicons
A crucial element of sentiment analysis is the **financial lexicon**, like the Dow Jones Lexicon (DJL), which comprises terminology and dictionaries designed for financial situations. For instance, when a news story includes terms like “dropped,” “lost,” or “fell,” the computer identifies these as negative indicators, thus lowering the sentiment score for the referenced stock.

Additionally, context matters. Words featured in headlines or initial paragraphs statistically have greater significance than those buried deep within the article. Algorithms account for this hierarchy to create a detailed sentiment score. This scored output has a direct impact on trading notifications, suggesting actions like selling, buying, or holding.

#### Behind the Scenes: Sentiment in Code
Financial algorithms rely on organized data formats like **XML (Extensible Markup Language)** to structure and handle text data. Unlike HTML, which organizes the visual design of a web page, XML organizes the data content itself. For example, a sentiment database structured in XML may resemble the following:

“`xml

MSFT
Bearish

AMZN
Bullish

“`

A sentiment system aggregates real-time news and market analysis into this structured format, enabling trading models to efficiently comprehend and respond to valuable insights.

While sentiment analysis offers glimpses into market mood, advanced methods in **machine learning** take it further—predictive modeling.

—

### Machine Learning: Anticipating Stock Market Trends

#### What is Machine Learning?
Machine learning (ML) is a branch of artificial intelligence (AI) that allows systems to learn and refine their capabilities based on available data. In stock market trading, ML models are designed to identify patterns within financial data and predict stock or portfolio movements.

#### The Training Process
To create an ML model, historical stock data (features such as **opening price**, **high/low of the day**, and **trading volume**) is input into the system. Consider the training data below for Microsoft stock in January 1990:

As raw data can be cumbersome, values are often **normalized** to conform within a 0 to 1 range, improving processing speed and precision. Following normalization, this table would be represented as follows: