trade-ahead-unsupervised-learning

Context

The stock market is widely recognized as a beneficial investment avenue for long-term savings and wealth creation. Investing in stocks offers multiple advantages, such as combating inflation, accumulating wealth, and enjoying certain tax benefits. Over time, steady returns on investments can significantly increase one's savings, often more than one might expect. Additionally, the power of compound interest means that starting investments early can result in a larger fund for retirement, helping to meet various financial goals throughout life.

However, it's crucial to maintain a diversified portfolio when investing in stocks to maximize earnings across different market conditions. A diversified portfolio not only has the potential for higher returns but also reduces risk by mitigating losses during market downturns. Navigating the complex world of financial metrics to assess stock value can be overwhelming, especially when evaluating numerous stocks to determine the best choices for an individual. Cluster analysis can simplify this process by grouping stocks with similar traits and identifying those with minimal correlation. This approach allows investors to effectively analyze stocks across various market segments, enhancing their ability to shield their portfolios from potential risks and losses.

Objective

Trade Ahead, a financial consultancy firm, offers tailored investment strategies to its clients. They've brought me on board as a Data Scientist and supplied me with data that includes stock prices and various financial indicators for several companies listed on the New York Stock Exchange. My assignment involves analyzing this data, categorizing the stocks according to the attributes given, and providing insights into the characteristics of each group.

EDA

1. How are stock prices distributed?
2. How much have stock prices increased by economic sector?
3. What correlations exist between different variables?
4. What is the cash ratio across different economic sectors?
5. What are the P/E ratios within each economic sector?

Data Dictionary

• Ticker Symbol : This is a unique abbreviation used to identify publicly traded shares of a particular stock on a stock market.
• Company : The name of the company.
• GICS Sector : The economic sector assigned to a company by the Global Industry Classification Standard (GICS), which reflects the primary business activities of the company.
• GICS Sub Industry : A more specific category within the GICS that pinpoints the company's business operations.
• Current Price : The current trading price of the stock, expressed in dollars.
• Price Change : The percentage increase or decrease in the stock price over the last 13 weeks.
• Volatility : The standard deviation of the stock price over the past 13 weeks, indicating how much the price fluctuates.
• ROE (Return on Equity) : This is calculated by dividing the company's net income by its shareholders' equity, which shows how effectively management is using a company’s assets to create profits.
• Cash Ratio : A measure of a company's liquidity, calculated as the ratio of its cash and cash equivalents to its current liabilities.
• Net Cash Flow : The net amount of cash and cash equivalents moving into and out of a business.
• Net Income : The company's total earnings, factoring in costs, expenses, interest, and taxes, expressed in dollars.
• Earnings Per Share : The portion of a company's profit allocated to each outstanding share of common stock, in dollars.
• Estimated Shares Outstanding : The total number of a company’s shares currently held by all its shareholders.
• P/E Ratio : The ratio of the company’s current stock price to its earnings per share, indicating how much investors are paying for a dollar of earnings.
• P/B Ratio : A ratio that compares a company's market value to its book value, calculated as the stock price per share divided by the book value per share (the net asset value of a company).

Skills

• Exploratory Data Analysis : Delving deep into datasets to uncover underlying patterns and insights.
• Data Preprocessing : Cleaning and preparing data to ensure it is analysis-ready.
• Data Scaling : Normalizing or standardizing data to ensure uniformity for analytical models.
• K-means Clustering : Implementing K-means to identify distinct groups within data based on similarity.
• Hierarchical Clustering : Applying hierarchical methods to build a tree of clusters and understand data hierarchy.
• Model Building (Unsupervised Learning) : Developing models to analyze data without pre-labeled outcomes, focusing on discovering hidden structures.
• Cluster Profiling : Characterizing and understanding the properties of each cluster to interpret model outputs and derive insights.

Insights

Analysis was performed on the dataset and 4 distict groupings of stocks were identified using k-means clustering.

Group 1 : High priced stocks with moderate volatility and very high earnings per share. These are very profitable/valuable companies with a high cash ratio that can potentially return 5-10x ROE.

Group 2 : Moderate priced stocks with low volatilty and moderate earnings per share. These are safe stocks that do not fluctuate very much but retain their value. Ideal for low risk investments.

Group 3 : Low priced stocks with high volatility and negative cash ratio. These stocks belong to startups, companies that are new and have yet to prove themselves but have high growth potential.

Group 4 : Moderate priced stocks with high volatility and high earnings per share. Price change can go both ways. These represent companies that are on the rise and are taking on investments to scale up.

Which clustering technique was quicker? Generally, the K-means algorithm required less time for execution as it is less computationally demanding compared to hierarchical clustering.

Which clustering technique produced more distinct clusters? The K-means algorithm typically formed 4-6 clear clusters, whereas the hierarchical clustering method yielded 5-8 distinct clusters.

How similar are the clusters from both techniques in terms of observations? Both methods identified a similar cluster characterized by stocks with exceptionally high current prices, moderate volatility, very high cash ratios, and a return on equity (ROE) ranging from 5 to 10 times.

What is the optimal number of clusters identified by each algorithm? For K-means: Utilizing the elbow method and silhouette scores, 4 clusters were determined to be optimal. For Hierarchical clustering: Using average linkage and Euclidean distance, which produced the highest cophenetic correlation, 5 clusters were identified as optimal.

Recommendations for Business

Based on the insights from the analysis of stock groupings using k-means clustering, here are some recommendations to help an investment business diversify their clients' portfolios:

Group 1: High-Value Stocks

Characteristics : High-priced, moderate volatility, very high earnings per share.

Why Choose : These stocks are from profitable and valuable companies, ideal for investors looking for high returns through stable, well-established companies.

Group 2: Low-Risk, Stable Stocks

Characteristics : Moderate-priced, low volatility, moderate earnings per share.

Why Choose : These stocks offer stability and are less affected by market fluctuations, making them suitable for risk-averse investors focused on capital preservation.

Group 3: High Growth Potential Stocks

Characteristics : Low-priced, high volatility, often negative cash ratio.

Why Choose : These stocks belong to startups or rapidly growing companies with potential for substantial returns. They are best for investors who are risk-tolerant and looking for aggressive growth.

Group 4: Emerging Companies

Characteristics : Moderate-priced, high volatility, high earnings per share.

Why Choose : These are stocks of companies in the growth phase, likely to scale up and increase in value. Suitable for investors looking for growth opportunities with a readiness to handle volatility.

General Allocation Principles:

• Diversify Across Groups : Ensure that investments are spread across different groups to mitigate risks and capitalize on various growth potentials.
• Tailor to Individual Risk Tolerance : Adjust the percentage allocation in each group based on the individual’s risk tolerance, investment goals, and time horizon.
• Regular Reassessment and Rebalancing : Market conditions and company performances change over time, necessitating periodic reassessment and rebalancing of the portfolio to maintain alignment with the investor’s objectives.