Accurate climate condition detection plays a crucial role in understanding long-term environmental changes and predicting future climate behavior. By analyzing variations in temperature, precipitation, and atmospheric trends, it becomes possible to identify global warming patterns and their regional impacts. This paper analyzes global and regional climate anomaly trends using traditional time-series and machine learning models, including Linear Regression, Ridge Regression, Random Forest, ARIMA, and Holt-Winters. The dataset, representing temperature anomalies relative to the 1951–1980 baseline, was used to forecast trends up to 2030. Results show a consistent rise in global temperatures across all models, confirming the persistent impact of climate change. The Holt-Winters model achieved the highest accuracy (MAE = 0.1868, RMSE = 0.2083, MAPE = 13.13%), effectively capturing long-term trends, while ARIMA also performed competitively. Random Forest excelled in capturing non-linear regional patterns, particularly for Australia, Brazil, and Germany, where MAPE values ranged from 15–26%. Overall, integrating statistical and machine learning approaches enhances forecasting accuracy and supports data-driven climate resilience planning.

Climate anomaly detection; Temperature forecasting; Machine learning; Linear Regression; Ridge Regression; Random Forest; ARIMA; Holt–Winters