busRidershipPrediction-RTA Dubai is a project that aims to predict the daily ridership at a given bus stop for a specific route in Dubai. By utilizing the DubaiPulse - RTA-BusRidership daily dataset, along with additional external data such as holiday information, weather data, and bus station data, the project leverages machine learning and deep learning techniques to provide accurate ridership predictions. XGBoost (XGB) and LightGBM (LGB) algorithms are fine-tuned and tested, alongside a dense layer from the TensorFlow/Keras libraries.
- The mean absolute error (MAE) metric is used to evaluate model performance, with achieved MAE ranging from 10 to 14 passengers across different models.
- Data Collection and Preprocessing: Aggregate and preprocess the daily bus ridership data from DubaiPulse - RTA-BusRidership dataset. Incorporate external data such as holiday information, weather data, and bus station details for enhanced predictions. Machine Learning Algorithms: Utilize XGBoost (XGB) and LightGBM (LGB) algorithms to build accurate ridership prediction models. Fine-tune the models to achieve optimal performance.
- Deep Learning with Dense Layers: Implement a deep learning approach using a dense layer from the TensorFlow/Keras libraries to further enhance prediction accuracy.
- Performance Evaluation: Measure the performance of the models using the mean absolute error (MAE) metric. Analyze and compare the MAE scores across different machine learning and deep learning models.
- Reference: https://www.scitepress.org/papers/2017/63457/pdf/index.html