https://github.com/ifigeneiatsiflidou/applied-statistics-project
Project for an Applied Statistics course, involving exploratory data analysis and predictive modeling of movie revenue using engineered features and multiple linear regression.
https://github.com/ifigeneiatsiflidou/applied-statistics-project
correlation-analysis data-analysis linear-regression python scikit-learn visualization
Last synced: about 2 months ago
JSON representation
Project for an Applied Statistics course, involving exploratory data analysis and predictive modeling of movie revenue using engineered features and multiple linear regression.
- Host: GitHub
- URL: https://github.com/ifigeneiatsiflidou/applied-statistics-project
- Owner: ifigeneiatsiflidou
- Created: 2025-04-14T14:39:39.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-04-14T17:10:21.000Z (about 1 year ago)
- Last Synced: 2025-04-15T04:16:00.471Z (about 1 year ago)
- Topics: correlation-analysis, data-analysis, linear-regression, python, scikit-learn, visualization
- Language: Jupyter Notebook
- Homepage:
- Size: 7.46 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Applied Statistics Project
This project was created as part of an **Applied Statistics** course and focuses on analyzing and predicting the worldwide box office revenue of movies using statistical techniques and data analysis in Python.
## Project Overview
The dataset includes 3,000 movies from **The Movie Database (TMDb)**, featuring details like: budget, popularity, language, cast, and production companies. The project is divided into 2 exercises:
## Exercises
### Exercise 1: Exploratory Analysis and Feature Selection Based on Correlation
We begin by analyzing the relationship between the revenue and four explanatory variables:
- budget
- english (a binary variable indicating whether the original language is English)
- runtime
- popularity
For each numerical explanatory variable, we compute the correlation coefficient with revenue and visualize the relationships using scatter plots.
Based on the correlation values and visual evidence, we determine which single variable is the most informative for predicting movie revenue.
### Exercise 2: Multiple Linear Regression
We extend our analysis by exploring additional variables that may improve our ability to predict movie revenue.
Most of the dataset’s remaining information is in text or JSON format. From this, we engineer two new numerical explanatory variables:
- The number of male cast members per movie (from the cast field)
- The number of production companies involved in each movie (from the production_companies field)
We then:
- Calculate the correlation of these new variables with revenue
- Select the four most predictive features (budget, popularity, men, companies)
- Fit a multiple linear regression model using scikit-learn
- Evaluate the model using the R² score and a histogram of residuals
- Interpret the model coefficients and discuss variable importance
The goal is to understand which variables most strongly affect movie revenue and how well a linear model can explain the variability in the data.
## Contents
`movie_revenues.csv` : Dataset with 3,000 movies
`Applied_statistics_project.ipynb` : Main Jupyter notebook with code and analysis
`Applied-Statistics-Project.pdf`: Final report