GitHub - amanvora/Movie-Rating-Prediction: A regression system was designed that predicts the IMDb rating of a movie. A movie viewer would otherwise have to rely on a critic's review or self-instincts. A dataset, obtained from Kaggle, contains certain attributes (such as genre, duration, names of actor, director, number of voters for the rating, plot and keywords, language, etc.) and performance rating of over 5000 movies. On observation, the dataset is preprocessed in order to fill missing data and convert categorical features to numerical. Several hypotheses are selected and a few not chosen due to unfavorable conditions observed. A 5-fold cross-validation on the hypotheses set exhaustively searches for the most optimum hyperparameters of the model with the least MSE and helps curtail overfitting. Random forest regressor is chosen due to its low MSE and simplicity. The model is retrained and gives a better coefficient of determination on the test set when compared to previous techniques on Kaggle. Software: Python with use of Scikit-learn library

amanvora / Movie-Rating-Prediction Public

Notifications You must be signed in to change notification settings
Fork 3
Star 3

A regression system was designed that predicts the IMDb rating of a movie. A movie viewer would otherwise have to rely on a critic's review or self-instincts. A dataset, obtained from Kaggle, contains certain attributes (such as genre, duration, names of actor, director, number of voters for the rating, plot and keywords, language, etc.) and per…

3 stars 3 forks Branches Tags Activity

Star

Notifications

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
README.txt		README.txt
modelSelCV.py		modelSelCV.py
modelTest.py		modelTest.py
movie_metadata.csv		movie_metadata.csv
preproc.py		preproc.py

Repository files navigation

Libraries required:
1. Scikit-learn
2. Matplotlib
3. numpy



Instructions to run:
1. Please run preproc.py first. It will generate 'preprocessed.txt' and 'imdbScores.txt' to be used by the other two scripts.

2. modelSelCV.py and preproc.py each have commented sections in order to view plot of covariance matrix and categorical features respectively.

3. Since model selection performs an exhaustive evaluation of all the models across all the hyperparamters, it takes a long time to execute (approx. 30 minutes depending on the system)




Code referred:
1. For one-hot enconding of catgorical features:
http://fastml.com/converting-categorical-data-into-numbers-with-pandas-and-scikit-learn/

2. For splitting the data to (train,test) and loading the database
https://www.kaggle.com/nsnilay/d/deepmatrix/imdb-5000-movie-dataset/predicting-imdb-ratings-using-numerical-attributes/notebook