Weatheria

Weatheria predicts Earth surface temperatures using advanced machine learning techniques. By analyzing historical climate data, it forecasts global warming trends based on spatial (latitude, longitude) and temporal (year, month) features. The insights aim to contribute to a deeper understanding of climate change's impact and help build resilience against its effects.

• Frontend: Built using Nuxt.js with TailwindCSS for styling and extended with DaisyUI components for a modern, responsive design. The frontend is deployed on Vercel.
• Backend: Hosted on Google Cloud's Virtual Machine (VM) and containerized using Docker for easy scalability and reliable deployment. The backend was developed with FastAPI, responsible for serving predictions from pre-trained machine learning models.
• Machine Learning: The models were built using Python and the Scikit-learn library. Four algorithms were used: Random Forest, K-Nearest Neighbor (KNN), Support Vector Regression (SVR), and Linear Regression. The Random Forest model achieved the highest accuracy with an R² score of 0.9857.
• Images and Illustrations: The visuals used in the project are sourced from Storyset, under their license agreement. Special thanks for providing high-quality illustrations to enhance the user interface.

The dataset used in this project is sourced from the Berkeley Earth Surface Temperature repository, a trusted resource for historical climate data. It includes records of average land temperatures collected globally, along with metadata such as date, latitude, longitude, and uncertainty levels. To ensure data quality:

• Records prior to 1870 were excluded due to data incompleteness.
• Missing temperature values were removed to maintain dataset integrity.
• Features such as year, month, latitude, and longitude were carefully extracted for model training.

These steps resulted in a clean and structured dataset, allowing the models to effectively capture spatial and temporal patterns in global temperature changes.

Four machine learning algorithms were employed to analyze the data: Random Forest, K-Nearest Neighbor (KNN), Support Vector Regression (SVR), and Linear Regression. Each model was evaluated using metrics such as Mean Absolute Error (MAE), Median Absolute Error (MedAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R² scores. Random Forest emerged as the most accurate model with an R² score of 0.9857.

The project analyzed global climate data to predict temperature trends with remarkable accuracy. Below is a summary of the performance of the models: