https://github.com/ikeasamoahansah/zindi-hack

A machine learning model to predict carbon emissions
https://github.com/ikeasamoahansah/zindi-hack

Last synced: 1 day ago
JSON representation

A machine learning model to predict carbon emissions

• Host: GitHub
• URL: https://github.com/ikeasamoahansah/zindi-hack
• Owner: ikeasamoahansah
• License: cc0-1.0
• Created: 2024-08-15T19:08:08.000Z (3 months ago)
• Default Branch: main
• Last Pushed: 2024-09-17T23:55:25.000Z (about 2 months ago)
• Last Synced: 2024-09-18T04:06:27.637Z (about 2 months ago)
• Language: Jupyter Notebook
• Size: 9.26 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
CO2 emission prediction challenge 🌨️ 


- The model was run as a Kaggle notebook in which datasets were saved

- Datasets: https://www.kaggle.com/datasets/ikeasamoahansah/co2-prediction-dataset/data

### 1. Setup

- Test and Train files could not be uploaded to GitHub due to their sizes

- Files can be found in the dataset link above

- Can be run straight from [kaggle](https://www.kaggle.com/code/ikeasamoahansah/test-features-emission-py)

- test and train variables should be replaced with their respective paths

### 2. Order

- The code should be run in the normal order (from top to bottom)

- The file to be used is the [final_submission.ipynb](final_submission.ipynb) file

### 3. Explanations of Features Used

- Missing data was imputed and Regressors were used to predict the missing values

- Log, Power and Other transformations were used to normalize the features in the training set.

- Other features were added based on location and rotated in degrees from some findings.

- Features were then gathered from all and grouped as top 50 best features

- These top 50 went further preprocessing and were trained

- Catboost and LGBM Regressors were used for final training along with some Ensemble methods

### 4. Environment

- The model was trained on kaggle with a GPU P100 by Nvidia (all free to use) can also be run on a CPU

### 5. Hardware

- A CPU or GPU is fine.

- GPU required for faster training time

### 6. Expected run time

- with a GPU: 3 hrs 45 mins

- only CPU: 8 hrs 30 mins