Review of Techniques and Algorithms of Temperature Prediction Using Artificial Intelligence

Abstract

This study investigates recent advancements in Artificial Intelligence (AI) techniques for temperature prediction, offering a structured review guided by five key research questions. The novelty of this work lies in its comprehensive analysis of 48 carefully selected research papers (from 2018 to 2024), screened through defined exclusion criteria, to identify dominant trends, effective methods, and future directions in AI-based weather forecasting. The results reveal that deep learning models were the most commonly applied techniques, appearing in 17 out of 48 manuscripts (35.41%). Regarding the focus of the studies, 35 papers (74.46%) employed specialized predictive algorithms tailored for temperature forecasting. Geographically, Asia was the leading region in contributions, with India alone accounting for 10 papers (20.83%). In terms of data sources, approximately 50% of the studies used sensor-based climate data, emphasizing the reliance on real-time environmental inputs. For veracity, Long Short-Term Memory (LSTM) networks and Deep Neural Networks (DNN) proved to be the most successful for time series predictions and on the other hand Random Forest (RF) and Support Vector Machines (SVM) have shown to be more appropriate for classification problems like comparative factor analysis. This article is organized systematically to offer the readers clear, concise and practical views about the utilization of AI for weather monitoring systems so that they could make use of AI in their sustainability projects.