This repository contains a machine learning project focused on forecasting absolute humidity in San Diego using a Supervised Hybrid Production-Level Long Short Term Memory (LSTM) Recurrent Neural Network. The project aims to predict next-day humidity based on current weather parameters, showcasing advanced techniques in time-series analysis and neural network optimization. This project was not completed but contains function that I have recycled for future neural networks optimization and helped with decision making during construction.
- Time Series Forecasting: Employing LSTM networks to model and predict weather patterns.
- Data Preprocessing: Handling and transforming real-world weather data for machine learning applications.
- Feature Selection: Utilizing techniques like Recursive Feature Elimination (RFE) and Sequential Feature Selection (SFS) to enhance model performance.
- Statistical Analysis: Applying Mean Absolute Error (MAE) and other statistical measures to evaluate model accuracy.
- Machine Learning Optimization: Fine-tuning and optimizing LSTM models for better prediction accuracy.
- Weather-Dependent Decision Making: Providing accurate humidity forecasts, crucial for sectors like agriculture, environmental management, and urban planning.
- Resource Optimization: Enabling businesses and organizations to plan more effectively by anticipating weather conditions.
- Data-Driven Insights: Offering valuable insights into weather patterns, potentially contributing to climate research and sustainability initiatives.
This project demonstrates my capability to handle complex time-series data and build sophisticated machine learning models. It reflects my skills in:
- Analyzing and interpreting large datasets.
- Developing and optimizing advanced predictive models.
- Applying machine learning techniques to solve real-world problems.