permanatayev / movie-revenue-prediction

Model that uses 10 different algorithms to predict the revenue of a movie before it's release

Movie-revenue-prediction

Introduction

The project aims to predict, with high accuracy, the success (revenue) of a movie before its release by using all information that we could have obtained before the movie’s release. Multiple models will be used and compared against each other to find the best one.

Problem Definition

Given information about a movie regarding its release, can an algorithm predict the total gross revenue that movie will generate? Such information would be useful to marketers, theater operators, and others in the movie industry, but it is a hard problem, even for human beings. Based on the dataset, which is The Movies Dataset, and the problem described above, we experimented with different models to predict the revenue. Apart from making an accurate prediction, we are interested in studying the effect of our data on the prediction, i.e. what features, engineered features and feature combinations have what effects. We are also interested in seeing how different types of models performed in the context of our problem.

Methods

For training we used the models below:

1. Artificial Neural Network
2. Light Gradient Boosting Machine
3. Random Forest
4. XGBoost
5. CatBoost Regressor
6. Elastic-Net
7. Ridge Regression
8. Linear Regression
9. Polynomial Regression
10. Support Vector Regression