This project focuses on predicting YouTube video revenue using different machine learning techniques, including Linear Regression, Artificial Neural Networks (ANN), and Support Vector Regression (SVR). The goal is to build accurate models that estimate revenue based on various input features.
- Data Collection and Preprocessing: We collected a diverse dataset of YouTube videos and performed thorough data cleaning and preprocessing.
- Feature Selection and Engineering: Relevant features were selected and engineered, including Date, Average_views_per_viewer, Unique_viewers, Impressions, Comments added, Shares, Likes_(vs._dislikes), etc.
- Linear Regression: A baseline linear regression model was built to establish initial insights into the data.
- Artificial Neural Networks (ANN): A deep learning ANN architecture was developed to capture complex relationships in the data.
- Support Vector Regression (SVR): SVR was used to handle nonlinear relationships between features and revenue.
- Model Evaluation and Comparison: Models were evaluated using metrics such as MAE, RMSE, and R-squared.
- Interpretability and Insights: Feature importance and coefficients were analyzed to gain insights into revenue influencers.
- Clone this repository:
git clone https://github.com/ankita-1007/youtube-revenue-prediction.git - Run the Jupyter Notebook files for data preprocessing, model training, and evaluation.
Achieved an accuracy of 96.19 using Linear Regression, 83.28 using ANN and 97.24 using SVR.
This project demonstrates the application of Linear Regression, ANN, and SVR models to predict YouTube video revenue. By leveraging machine learning techniques, content creators can gain insights into revenue optimization strategies for their videos.