This project focuses on predicting future weather conditions—specifically temperature values—using a Long Short-Term Memory (LSTM) deep learning model. Weather forecasting plays an essential role in agriculture, transportation, disaster management, and daily decision-making. Traditional forecasting methods can struggle with complex, non-linear climate patterns, whereas LSTMs are excellent at understanding time-series data like weather trends.
To build and evaluate an LSTM-based neural network that can learn patterns from historical weather data and accurately forecast future temperature values.
- Uses historical weather dataset for training and testing
- Data preprocessing includes normalization, handling missing values, and feature scaling
- Time-series split into sequences for LSTM input format
- Model built using TensorFlow / Keras
- Training and validation used to tune prediction performance
- Results visualized through prediction vs actual graphs
- LSTM (Long Short-Term Memory) Neural Network
- Optimized for sequential data and long-term dependencies
- Predicts future temperature values based on learned trends
- Visual comparison chart showing performance accuracy
The LSTM model demonstrated high capability in capturing temporal weather patterns and produced reliable forecasts. With more features such as humidity, wind speed, and pressure—and a larger dataset—the prediction accuracy could improve even further. This approach can be extended into real-world weather prediction systems.