Scholarlink

Introduction: This study analyzes long-term relative humidity (RH) data from the Des Moines Station (DSM), Iowa, USA, from 1936 to 2024, using traditional and artificial intelligence (AI) approaches to improve forecasting accuracy. Understanding RH trends using AI is crucial for climate science, agriculture, and public health. Objectives: The research aims to predict long-term RH changes using both conventional statistical and advanced AI techniques. Key objectives include preprocessing daily RH data into a monthly time series format; applying ARIMA as a baseline model for linear trend forecasting; implementing deep learning models, LSTM and Bidirectional LSTM (Bi-LSTM), for nonlinear pattern analysis; using CNN to identify short-term variations though pattern recognition and developing hybrid models (CNN-LSTM and CNN-BiLSTM) for enhanced accuracy; and evaluating model performance using MAE, RMSE, MAPE, and R-squared metrics to identify the most effective model for predicting RH trends. Methodology: The study converted daily RH data into a monthly series, using ARIMA as a baseline. LSTM, Bi-LSTM, CNN, and hybrid CNN-LSTM models were applied to capture both short-term and long-term patterns. Model accuracy was evaluated using MAE, RMSE, MAPE, and R-squared. Findings: Results showed that hybrid AI techniques CNN-LSTM and CNN-Bi-LSTM models outperformed ARIMA, accurately capturing RH trends. This approach offers valuable insights for environmental monitoring and sectors sensitive to humidity changes.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Statistical AI Models for Environmental Sustainability: ARIMA, LSTM, and CNN-LSTM in Climate Prediction

  • 章节

  • 发表日期：

    2025年

查看全文
  • Serifat A. Folorunso,
  • Saheed A. Afolabi,
  • Olawale Dele Osanyintupin,
  • Morufu Aderemi Folorunso,
  • Richard O. Kehinde,
  • Aliu Adeniran Adebiyi,
  • Abdulsamad Adeola Folorunso,
  • Sukurat Salam,
  • Timothy Tomiwa Oyelakin

摘要

Introduction: This study analyzes long-term relative humidity (RH) data from the Des Moines Station (DSM), Iowa, USA, from 1936 to 2024, using traditional and artificial intelligence (AI) approaches to improve forecasting accuracy. Understanding RH trends using AI is crucial for climate science, agriculture, and public health. Objectives: The research aims to predict long-term RH changes using both conventional statistical and advanced AI techniques. Key objectives include preprocessing daily RH data into a monthly time series format; applying ARIMA as a baseline model for linear trend forecasting; implementing deep learning models, LSTM and Bidirectional LSTM (Bi-LSTM), for nonlinear pattern analysis; using CNN to identify short-term variations though pattern recognition and developing hybrid models (CNN-LSTM and CNN-BiLSTM) for enhanced accuracy; and evaluating model performance using MAE, RMSE, MAPE, and R-squared metrics to identify the most effective model for predicting RH trends. Methodology: The study converted daily RH data into a monthly series, using ARIMA as a baseline. LSTM, Bi-LSTM, CNN, and hybrid CNN-LSTM models were applied to capture both short-term and long-term patterns. Model accuracy was evaluated using MAE, RMSE, MAPE, and R-squared. Findings: Results showed that hybrid AI techniques CNN-LSTM and CNN-Bi-LSTM models outperformed ARIMA, accurately capturing RH trends. This approach offers valuable insights for environmental monitoring and sectors sensitive to humidity changes.