Pre-release Movie Rating Distribution Predict

This project is a extremely complete implementation of paper 《Pre-release Prediction of Crowd Opinion on Movies by Label Distribution Learning. IJCAI. 2015》.

Data

All training and testing Data are crawled from IMDb.com, reference to ldsvr/oldmoviecrawler.py and ldsvr/newmoviecrawler.py for details.

How can I crawl movie data from IMDb? Just reference to:

• IMDB电影数据库
• IMDB数据库结构

Feature Processing and Selection

Having training movies and testing movies, we need to process and select features from these data, as the paper said.

Reference to ldsvr/features_counting.py, ldsvr/features_indexing.py, ldsvr/matrix_persistent.py, ldsvr/test_matrix.py and ldsvr/training_matrix.py for details. And I use Redis DB for convenient feature counting and processing.

This process can form training matrix and testing matrix for LDSVR Algorithm.

Train

Following the algorithm from paper, I use LDSVR Algorithm to train the model. Please reference to ldsvr/ldsvr.py and paper for details.

Results

I use Tornado to show results. Here are some website snapshots shown below. And the website need mysql DB for data support.

References

• Geng, Xin, and Peng Hou. "Pre-release Prediction of Crowd Opinion on Movies by Label Distribution Learning." IJCAI. 2015.
• IMDb
• Tornado
• Redis
• Readfree.me