[Paper Review] Accurate Prediction of Temperature Indicators in Eastern China Using a Multi-Scale CNN-LSTM-Attention model

Jiajiang Shen, Weiyan Wu+1 more|arXiv (Cornell University)|Dec 11, 2024

Hydrological Forecasting Using AI5 citations

TL;DR

A multi-scale CNN-LSTM-attention architecture for time-series temperature Forecast in Eastern China, achieving MSE of 1.978295 and RMSE of 0.8106562.

ABSTRACT

In recent years, the importance of accurate weather forecasting has become increasingly prominent due to the impacts of global climate change and the rapid development of data science. Traditional forecasting methods often struggle to handle the complexity and nonlinearity inherent in climate data. To address these challenges, we propose a weather prediction model based on a multi-scale convolutional CNN-LSTM-Attention architecture, specifically designed for time series forecasting of temperature data in China. The model integrates Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and attention mechanisms to leverage the strengths of spatial feature extraction, temporal sequence modeling, and the ability to focus on important features. The development process of the model includes data collection, preprocessing, feature extraction, and model building. Experimental results show that the model performs excellently in predicting temperature trends with high accuracy. The final computed results indicate that the Mean Squared Error (MSE) is 1.978295 and the Root Mean Squared Error (RMSE) is 0.8106562. This work marks a significant advancement in applying deep learning techniques to meteorological data, offering a valuable tool for improving weather forecasting accuracy and providing essential support for decision-making in areas such as urban planning, agriculture, and energy management.

Motivation & Objective

• Address the need for accurate weather forecasting amidst nonlinearity and complex climate data.
• Develop a deep learning model that integrates spatial feature extraction, temporal sequence modeling, and attention to highlight important features.
• Apply the model to temperature time-series data in China to improve forecasting accuracy.
• Demonstrate the potential of deep learning for urban planning, agriculture, and energy management via better weather predictions.

Proposed method

• Construct a multi-scale CNN-LSTM-Attention neural architecture.
• Combine CNN for spatial feature extraction with LSTM for temporal dependencies.
• Incorporate an attention mechanism to focus on salient features.
• Preprocess data and perform feature extraction and model training.
• Evaluate using standard regression metrics including MSE and RMSE.

Experimental results

Research questions

• RQ1 Can a multi-scale CNN-LSTM-Attention model improve temperature time-series forecasting for Eastern China compared to baseline methods?
• RQ2 What is the impact of multi-scale CNN components and attention on predictive accuracy for climate indicators?
• RQ3 How do MSE and RMSE values reflect the model's performance in predicting temperature trends in Eastern China?

Key findings

• The model achieves MSE of 1.978295.
• The model achieves RMSE of 0.8106562.
• The approach demonstrates high accuracy in predicting temperature trends for Eastern China.
• The work demonstrates the applicability of deep learning to meteorological data for decision support in urban planning, agriculture, and energy management.