This application is a machine learning model that given a set of descriptive labels will attempt to predict the rating of a movie with those attributes.
The program is designed to predict movie ratings using a Naive Bayes model.
The model was trained on the IMDB 5000 Movie Dataset available, which can be found
here, and computes results by
measuring the frequency of each label. For instance, the program would measure
the probability of an actor being in a movie (given by actor_(.*)_name) given
its rating. For numerical quantities (continuous variables), the program uses
a normal distribution to measure the probability density of a given variable.
For instance, the duration variable, which measures the length of a
film, assumes that the length is normally distributed.
The current implementation uses Numpy for computations on lists, as well as Matplotlib for displaying statistical charts. These can be installed via
pip install numpy matplotlib
To use the program, one can run:
python class.py <csv_file> (<label> <value>)*
Where (<label> <value>)* represents an arbitrary number of uses of label-value
pairs. For instance, to predict based on duration and director name, one would run:
python class.py ./movie_metadata.csv duration 162 director_name "Christopher Nolan"
As of the latest test, this result should yield the same rating as was given to "The Dark Knight Rises", which was a 162 minute movie directed by Christopher Nolan.