As a Data Scientist, my latest project involved developing a comprehensive video game recommendation system for Steam using MLOps principles. The project was a part of my learning journey in Henry’s Bootcamp, where I applied ETL, EDA, and machine learning techniques to real-world data.
The primary goal was to build an API for sentiment analysis and game recommendations. The API offers various functionalities, including recommending games similar to a given one, identifying the top and least recommended games per year, and analyzing user sentiments based on their reviews.
I employed Python, Pandas, Scikit-Learn, FastAPI, and Uvicorn to create a robust backend capable of handling large datasets and delivering real-time insights. One of the core features was a sentiment analysis model that categorized user reviews into positive, neutral, or negative sentiments. Additionally, I used cosine similarity to recommend games that users are likely to enjoy based on their past choices.
The most challenging aspect of the project was optimizing the machine learning models for speed and accuracy, which was crucial given the API’s real-time nature. This project not only allowed me to deepen my understanding of MLOps but also demonstrated the power of data science in enhancing user experiences on platforms like Steam.
By deploying the API on Render, I ensured it was accessible for testing and further improvements. This project reflects my commitment to using data-driven insights to solve complex problems, and it is a testament to the practical applications of data science in the gaming industry.