Explore the phases of developing a machine learning model for predicting outdoor air temperature. This repo includes data preprocessing, model exploration, optimization, and deployment stages using Python and ML libraries.

This repository houses a comprehensive machine learning project aimed at enhancing the accuracy of air temperature forecasts, which is crucial for sectors like energy generation, agriculture, and disaster management. The project is structured into five distinct phases, each focused on a specific aspect of machine learning development from data handling to deployment.

Motivation: Enhancing the reliability of weather predictions to improve safety, energy management, and industrial operations.

Problem Specification: Develop a predictive model for air temperature using diverse meteorological parameters.

Literature Review: Review of methodologies ranging from traditional numerical models to advanced machine learning techniques.

Dataset of Choice: Utilizing the "Climate Weather Surface of Brazil - Hourly" dataset with over 8.39 million samples.

Data Cleaning: Handling missing values, removing irrelevant features.

Feature Analysis: Examining feature correlation and importance.

Experimental Setup: Configuration for model training/testing, parameter tuning.

Model Testing: Evaluation of various models using performance metrics.

Advanced Modeling: Refinement using grid search and custom loss functions.

Validation: Use of k-fold cross-validation to ensure model effectiveness.

Final Model Selection: Decision Trees chosen for their performance and suitability.

Deployment: Development of a utility application for real-time temperature prediction.

Data Handling: Python, Pandas, NumPy

Machine Learning: Scikit-Learn, TensorFlow