Welcome to my site

Unlocking INSIGHTS
Innovating for Tomorrow
Quest for

Knowledge
Power of

Analytics
Building

Solutions
Navigating Complexity
Transforming Data

Into Action

M2M Tech - Data Talent Program

Mission: upskill driven professionals with a passion for Data Science and Machine Learning, connect them with employers across various sectors across Canada.

Portfolio Profiling Project Description: I developed a stock analysis tool that integrates financial metrics and interactive visualizations.

Data Collection:

I implemented functionality to download stock data and financial metrics for specified tickers using 'yfinance' and calculate yearly Sharpe ratios, designed a feature to calculate and visualize cumulative returns with interactive Plotly graphs, and created an interactive candlestick chart for each ticker with customizable time intervals.

Visulazations

I developed a heatmap for exploratory data analysis (EDA) and performed statistical analysis of Sharpe ratios across various tickers and years, incorporated data handling techniques such as NaN management, Min-Max scaling, and the calculation of a decision score for each ticker based on value, growth, and profitability metrics, and designed an interactive investment donut chart to assist in portfolio decision-making based on the calculated decision score.

Optimizing Stock Trading Decisions

In today's fast-paced financial markets, making informed stock trading decisions is crucial for maximizing returns and minimizing risks. This project aims to enhance trading strategies by utilizing agent-based models and analyzing historical stock data. The goal is to predict optimal buy and sell actions for stocks, helping traders make better decisions.

Reasoning

Agent-based modeling simulates the actions and interactions of individual traders (agents) in the market. By incorporating historical data, we can create realistic trading scenarios that reflect past market behavior. Our project explores the performance of two different trading agents, each with unique strategies for making buy and sell decisions based on historical stock price movements. The strategies employed by the agents were designed to capitalize on market trends, reacting to price changes and volume data. By comparing the performance of these agents, we aim to identify which strategies yield the best results in terms of portfolio growth over time.

Results

Our analysis produced several key findings, which are visually represented in the project 1 code section above. Results demonstrate the effectiveness of the agents' strategies in managing a stock portfolio formed by 10 different firms. The results of our project indicate that agent-based models can significantly enhance stock trading strategies. By simulating various trading actions and analyzing historical data, we identified effective buy and sell strategies that lead to improved portfolio performance. These insights provide valuable guidance for traders looking to optimize their stock trading decisions in real-world markets.

Research Process

Research Question

What are the effects of decision-making speed and risk tolerance on the stock trading strategies of two agents: one who prioritizes cautious, slow decision-making and minimizes stock sales during low-risk periods, versus another who opts for rapid transactions and liquidates all holdings in response to expected negative returns?

Agents

Agent-based modelling (ABM) in stock markets is a computational approach that simulates the interactions of individual agents (traders) to understand complex market dynamics. Key features of our ABM in this project include: heterogeneity, adaptation, firm based wait time, firm based expected return thresholds, weighted buy-sell transactions.

Data Collection

To gather stock data, we utilized Yahoo Finance for ten different stock symbols through Python. Following the download of historical market data, we calculated short-term and long-term moving averages to inform the agents' decision-making processes regarding buy or sell signals.

Analysis

We optimized expected return thresholds through several methods: first, by applying a consistent unique constant across all symbols; second, by deriving firm-specific values directly from the data. Lastly, we combined these scores after normalizing the data into a formula, represented as 𝑎∗𝑥+𝑏 to enhance decision-making processes. Solved the equation to maximize the agents final portfolio values.

Flexible Agents

Challenges

Adaptive agents do not have restrictions on the number of transactions they can execute. Additionally, there is no minimum threshold for buy and sell operations. This lack of limitations can lead to suboptimal strategies for human investors, as executing transactions in small quantities may not be ideal in a context where each transaction incurs costs.

Flexible Agents

Flexible agents employ a maximum of three distinct buy operations per day, targeting firms with the highest expected returns. Flexible agents retain all the essential characteristics of the earlier adaptive agents.

Results

The maximum final portfolio values for a $10,000 investment over 3 years were $28,555 for the slow agent and $41,750 for the fast agent, resulting in a performance improvement of approximately 46.2% for the fast agent. When comparing Adaptive and Flexible agents, it is evident that both achieved similar outcomes. Notably, the Flexible Slow Agent delivered superior results with fewer transactions. A similar grid search approach could be employed to identify the optimal final portfolio value for the equation 𝑎∗𝑥+𝑏 .

ZelihaUral Merpez

Unlocking INSIGHTS

Innovating for Tomorrow

Quest for

Knowledge

Power of

Analytics

Building

Solutions

Navigating Complexity

Transforming Data

Into Action

M2M Tech - Data Talent Program

Data Collection:

Visulazations

Optimizing Stock Trading Decisions

Reasoning

Results

Research Process

Research Question

Agents

Data Collection

Analysis

Flexible Agents

Challenges

Flexible Agents

Results

ChatGPT - Quotes

Thanks

Contact me