کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

A Comparative Analysis of Machine Learning Algorithms for Modeling Global Cinematic Taste Cultures in Iran

نویسندگان :

Erfan Shariatkhah¹ Mohammad Mahdi Koohjani Gooji² Mehrshad Khosraviani³

1- دانشگاه آزاد اسلامی واحد پرند 2- دانشگاه آزاد اسلامی واحد پرند 3- دانشگاه آزاد اسلامی واحد پرند

کلمات کلیدی :

Web scraping،machine learning classification،movie popularity،cinematic data mining

چکیده :

This study proposes and implements a web scraping framework designed to collect and analyze a corpus of over 60,000 film records from the 30nama and Filmazon platforms. Utilizing Python libraries, including BeautifulSoup and Scrapy, the framework extracts cinematic attributes—such as title, genre, director, cast, release year, rating, and viewership metrics. The raw data are subsequently cleaned and structured into a tabular format using the Pandas library for downstream analysis. Comparative analysis of the employed five machine learning models, namely Logistic Regression, Decision Tree, Random Forest, XGBoost, and Gradient Boosting, revealed that Logistic Regression demonstrated superior performance with an accuracy of 0.912, in contrast to Gradient Boosting, which showed the lowest accuracy at 0.908. Analysis of predictor significance demonstrated that cinematic popularity within the Iranian context is most substantially influenced by temporal proximity (release year) and perceived global reception, quantified by the number of votes on IMDb. These results offer actionable insights for content creators and recommendation systems in Iranian cinema.