https://github.com/aravinda-1402/flight_delay_prediction_using_machine_learning

This research investigates flight delay trends, examining departure time, airline, and airport factors. Regression machine learning meth- ods are utilized to predict delay contributions from various sources. Time-series models, including LSTM, Hybrid LSTM, and Bi-LSTM, are compared with baseline regression models such as Multiple Regression, Decisi
https://github.com/aravinda-1402/flight_delay_prediction_using_machine_learning

bilstm delay lstm lstm-neural-networks prediction regression timeseries-forecasting

Last synced: about 1 month ago
JSON representation

This research investigates flight delay trends, examining departure time, airline, and airport factors. Regression machine learning meth- ods are utilized to predict delay contributions from various sources. Time-series models, including LSTM, Hybrid LSTM, and Bi-LSTM, are compared with baseline regression models such as Multiple Regression, Decisi

• Host: GitHub
• URL: https://github.com/aravinda-1402/flight_delay_prediction_using_machine_learning
• Owner: aravinda-1402
• License: apache-2.0
• Created: 2024-04-20T23:47:59.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-08-08T14:09:17.000Z (about 1 year ago)
• Last Synced: 2025-03-12T16:17:52.309Z (7 months ago)
• Topics: bilstm, delay, lstm, lstm-neural-networks, prediction, regression, timeseries-forecasting
• Language: Jupyter Notebook
• Homepage:
• Size: 6.56 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Deciphering Air Travel Disruptions: A Machine Learning Approach ✈️

📄 **Paper link:** https://doi.org/10.48550/arXiv.2408.02802

## Conda Environment 📦

Provided as `conda_environment.yml`. Create using: 

```bash

conda env create -f conda_environment.yml

```

(source: [Managing Conda Environments](https://conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html))

⚠️ **NOTE**: There may be differences in models produced with this environment as the models reported were trained in Google Colab. The reported time-series models can be found under `models` as `bilstm_model.h5`, `lstm_model.h5`, and `hybrid_model.h5`.

## Notebooks 📓

### `src/preprocessing.ipynb`

Notebook for preprocessing.

Dataset sourced from: [Flight Delay and Cancellation Dataset 2019-2023](https://www.kaggle.com/datasets/patrickzel/flight-delay-and-cancellation-dataset-2019-2023?select=flights_sample_3m.csv)

Creates processed label-encoded data and processed one-hot-encoded data. Splits data into train and test sets. Saves the data in CSV format. One-hot-encoded data was experimented with but ultimately not chosen due lacking computational power to handle a massive increase in features.

### `src/baseline_models.ipynb`

Trains and evaluates baseline regression models: Linear Regression, Decision Tree, Random Forest, Neural Network, XG Boost.

### `src/build_lstm_data.ipynb`

Creates datasets for time series models:

1. Combines `X_train` and `X_test`

2. Sorts by date-time

3. Temporally splits the data

This is necessary because datasets used for the baseline models are shuffled.

### `src/time_series_models.ipynb`

Trains the time-series models.

### `src/evaluate_time_series_models.ipynb`

Evaluates the time-series models.