DataGuti/Kaggle-Competition---Predicting-CO2-emissions-with-Gradient-Boosting-Regression-Model

A short Kaggle competition with the goal of creating a Machine Learning predictive model that predicts Rwanda's CO2 emission levels based on satellites' information. # Kaggle-Competition---Predicting-CO2-emissions-with-Gradient-Boosting-Regression-Model A short Kaggle competition with the goal of creating a **Machine Learning predictive model** that **predicts Rwanda's CO2 emission levels based on satellites' information**. For this competition I had **only 24 hours** (since I discovered the last day) so I decided to build a single Machine Learning model and improve it as much as possible in the little time I had. On this occasion, there'll be only one jupyter notebook and the EDA will be very brief: it's only purpose will be choosing the right data to train the model. **Important note:** the **resulting model** and **training/test datasets** were **too big to upload** here, even being compressed. ## Introducing the case Tracking carbon emissions is fundamental when it comes to deciding on measures against climate change. For this challenge, the objective was to create a predictive model (using Machine Learning) trained on open-source CO2 emissions **data from Sentinel-5P sattelite** observations. The training data ranged from 2019 to 2021, and the **goal** was to **predict 2022 carbon emissions** based on it. Approximately 497 unique locations were selected from multiple areas in Rwanda. *For more detailed information on this competition, click the following link:* kaggle.com ## Competition's Rules Each person/team would have **unlimited submissions** (although there was a daily limit) and would have **around 3 weeks** to create a model. We could decide to not use some of the information provided in the dataset, *however*, we were **not allowed** to use any **external data**. The **evaluation metric** for this competition was the **RMSE** (Root Mean Square Error). ## Dataset available The dataset had **79023 entries** and **76 columns**, but the **main features** (without breaking them down) are: - ID (consists on a label made from the Latitude, Longitude, Year and …