Movie Metadata and Performance Dataset for Machine Learning Applications
Description
This dataset contains 5,674 unique movie records and 32 attributes prepared for machine learning, data analysis, and entertainment-domain research. It includes information such as movie title, release year, runtime, rating, vote count, Metascore, gross revenue, genre, certification, director, cast, and textual descriptions. The purpose of this dataset is to support the investigation of relationships between movie characteristics and performance-related indicators. The dataset enables exploration of questions such as how factors including genre, runtime, release year, audience voting behaviour, critical scores, and financial performance are related. It can also be used to investigate whether combinations of movie metadata features can support predictive or classification-based machine learning tasks. The data was prepared from real-world movie metadata through systematic data cleaning, transformation, missing-value treatment, text preprocessing, feature engineering, and statistical standardization. Redundant index artifacts were removed and duplicate records were checked. Missing numerical values were handled using genre-wise median imputation, while missing certification values were treated using categorical imputation. Cleaned and structured versions of genre, director, cast, and description fields were created, along with a primary genre feature. Additional standardized features were generated for movie rating, vote count, Metascore, gross revenue, runtime, and release year using Z-score and T-score transformations. These transformations allow numerical variables with different scales to be more easily compared and analysed. The dataset can be interpreted through exploratory statistical analysis, visualization, correlation analysis, regression, classification, clustering, and feature-based machine learning models. Potential applications include movie rating prediction, genre analysis, popularity and audience engagement analysis, financial performance analysis, recommendation system research, and natural language processing of movie descriptions. The final dataset is intended as a structured research resource for analysing patterns and relationships within movie metadata and for evaluating machine learning approaches using real-world entertainment-domain data.
Files
Steps to reproduce
1. Download or obtain the original real-world movie metadata source referenced in the accompanying dataset documentation. 2. Inspect the raw data structure and retain the relevant movie metadata attributes, including title, release year, runtime, rating, vote count, Metascore, gross revenue, genre certification, director, cast, and description. 3. Remove redundant index artifacts and inspect the dataset for duplicate records, missing values, inconsistent formatting, and invalid data types. 4. Clean text and categorical fields by converting stringified list-style values into structured, readable text. Create cleaned versions of genre, director, cast, and description attributes. 5. Extract the first available genre to create the Primary_Genre feature. 6. Handle missing numerical values by applying genre-wise median imputation to Metascore and Gross values. Handle missing certification values using categorical imputation. 7. Generate cleaned and engineered features, including Primary_Genre, cleaned text attributes, and imputed numerical attributes. 8. Standardize the numerical variables Movie Rating, Votes, MetaScore, Gross, Runtime, and Year of Release using Z-score transformation: Z = (X - μ) / σ 9. Generate corresponding T-score features using: T = 50 + 10Z 10. Perform final validation by checking the number of records, feature count, data types, remaining missing values, duplicate records, and numerical ranges. 11. Export the validated dataset as movie_dataset_final.csv.
Institutions
- Bharati Vidyapeeth Deemed UniversityMaharashtra, Pune