📊 Data Science Portfolio

Welcome to my project portfolio. Here, I showcase solutions developed to solve complex, real-world problems, covering the entire data lifecycle: from real-time data extraction and engineering to predictive modeling using Deep Learning.

My goal is to transform raw, unstructured data streams into automated strategic insights.

---

📂 Featured Projects

1. 📈 Stock Price Prediction using LSTM Neural Networks

Domain: Deep Learning, Time Series Forecasting, Quantitative Finance

This project develops a high-precision predictive model for financial time series, focusing on asset price volatility and closing trends in the stock market.

• The Problem: Stock prices possess complex temporal dependencies and short-term noise, rendering traditional regression models inefficient.
• The Solution: Implementation of a Recurrent Neural Network (RNN) using LSTM (Long Short-Term Memory) architecture, capable of retaining long-term patterns in time series without suffering from the vanishing gradient problem.
• Technical Approach:
  • Preprocessing: Historical data cleaning and scale normalization using Scikit-Learn's MinMaxScaler.
  • Modeling: Architecture built with TensorFlow/Keras, stacking LSTM layers with regularization mechanisms (Dropout) to mitigate overfitting.
  • Validation: Evaluation based on rigorous error metrics, such as RMSE (Root Mean Squared Error) and MAE (Mean Absolute Error).
• Core Technologies: TensorFlow, Keras, Scikit-Learn, Pandas, Matplotlib.
• Where to Find: /Stock_Price_Prediction_LSTM folder

---

📊 2. Geopolitical & Economic Intelligence Monitor (Truth Social)

Domain: Data Engineering, Real-Time Streaming, Advanced Web Scraping

An automated pipeline for real-time data extraction and monitoring targeting official profiles of public figures on the Truth Social platform. The system monitors mentions of strategic terms related to the market, oil, and geopolitical tensions.

• The Problem: Modern platforms utilize SPA ecosystems (React/Next.js) and strict firewalls (Cloudflare), blocking traditional scrapers (like BeautifulSoup) or causing loading deadlocks.
• The Solution: An asynchronous bot programmed in Stealth mode that, instead of parsing the visual HTML layout, intercepts raw JSON API packets within the browser network traffic.
• Technical Approach:
  • Stealth Mode: Script injection to mask automation variables (navigator.webdriver) and emulate human behavior.
  • Optimization: Dynamic blocking of media assets (images/fonts) to speed up synchronization by 82% and save processing power.
  • String Robustness: Structural handling of URLs via ASCII tables to safeguard the script against local compiler text encoding failures.
• Core Technologies: Playwright (Async API), Asyncio, Regular Expressions (re), Urllib.
• Where to Find: /Web_Scapping_TruthSocial folder

---

🛠️ Core Tech Stack

• Programming Languages: Python (Pandas, NumPy, SciPy)
• Artificial Intelligence & Deep Learning: TensorFlow, Keras, Scikit-Learn
• Data Engineering & Automation: Playwright, Asyncio, Requests, RESTful APIs
• Environments & Tools: PyCharm, Git, Virtual Environments (.venv)

---

📈 Key Outcomes & Skills Demonstrated

1. Time Series Expertise: Ability to prepare the three-dimensional matrices required by Deep Learning models ([samples, time steps, features]) and fine-tune neural network hyperparameters for financial forecasting.
2. Resilient Data Engineering: Development of scripts built to handle network failures, server fluctuations, and automated anti-bot blocks on web platforms.
3. Clean & Asynchronous Code: Advanced application of event-driven and asynchronous programming (async/await) in Python for high-performance optimization.