Podcast Listening Time Prediction

This project aims to predict the listening time (in minutes) of podcast episodes using machine learning techniques. The goal is to develop a model that accurately predicts how much time users spend listening to a particular podcast based on several features of the podcast episode.

Table of Contents

• Overview
• Dataset
• Model
• Installation
• Usage
• Results
• License

Overview

The project uses a Random Forest Regressor model to predict podcast listening times. The data consists of multiple features such as episode name, genre, and sentiment, along with other attributes that might influence how long a listener stays engaged with the podcast episode.

The model is evaluated using Root Mean Squared Error (RMSE), with the aim to minimize prediction error.

Dataset

The dataset used in this project is synthetic and contains information about various podcast episodes. It includes the following features:

• Episode: A unique identifier for each episode.
• Genre: The category or genre of the podcast (e.g., Comedy, Education, Technology).
• Sentiment: Sentiment associated with the podcast episode (e.g., Positive, Negative, Neutral).
• Listening_Time_minutes: The target variable, representing the time (in minutes) spent listening to the episode.
• Other numerical and categorical features related to the podcast.

Data Files

• train.csv: The training dataset containing the features and the target variable (Listening_Time_minutes).
• test.csv: The test dataset for making predictions (without the target variable).
• sample_submission.csv: A sample submission file in the required format for Kaggle.

Model

The machine learning model used for this project is a Random Forest Regressor, which is an ensemble learning method suitable for regression tasks. It was chosen for its ability to handle both numerical and categorical data effectively without requiring feature scaling.

Model Training

• The data is split into training (80%) and validation (20%) sets.
• The Random Forest model is trained on the training data and evaluated on the validation set using RMSE as the performance metric.
• The model is then used to predict the listening time on the test dataset.

Installation

To run the project on your local machine, follow these steps:

1. Clone the repository:

   git clone https://github.com/yourusername/podcast-listening-time-prediction.git
   cd podcast-listening-time-prediction

2. Set up a virtual environment:

   python -m venv myenv
   source myenv/bin/activate  # On Windows use 'myenv\Scripts\activate'

3. Install the required dependencies:

   pip install -r requirements.txt

Usage

1. Prepare the dataset: Place the train.csv and test.csv files in the data/ directory.

2. Train the model:

   python train.py

3. Predict listening times on the test data:

   python predict.py

4. Evaluate model performance on the validation set: After training, the model's RMSE on the validation set will be printed.

5. Submit the predictions: The predictions will be saved in a CSV file in the required format for Kaggle submission.

Results

The Random Forest model achieves an RMSE of 10.08 on the validation set, indicating that, on average, the predictions are off by about 10 minutes from the true listening times.

Next Steps

• Hyperparameter Tuning: The model can be fine-tuned using techniques like grid search or random search to improve its performance.
• Feature Engineering: Additional features or transformations could be added to further enhance model accuracy.
• Alternative Models: Exploring other regression models such as XGBoost or Gradient Boosting may yield better results.

License

This project is licensed under the MIT License - see the LICENSE file for details.