Understanding the spatiotemporal variability of precipitation is critical for effective water resource planning, particularly in regions with diverse climatic zones such as China. This study presents a hybrid methodology combining the Maximal Overlap Discrete Wavelet Transform (MODWT) and the Growing Neural Gas (GNG) clustering algorithm to regionalize precipitation patterns using monthly data from 123 synoptic stations over a 45-year period (1980–2024). MODWT was applied to decompose the precipitation time series into five frequency-based sub-series (W1–W5 and V5), capturing variability across 2- to 32-month cycles. Shannon entropy was calculated for each sub-series, generating a comprehensive feature set that reflects the temporal complexity at each station. These entropy features were subsequently used as input for the GNG algorithm, which identified 12 homogeneous precipitation clusters. The clustering performance was quantitatively assessed using the silhouette coefficient (SC), where the proposed model achieved a maximum SC value of 0.68, indicating strong inter-cluster separation and intra-cluster compactness. In comparison, clustering performed without MODWT-based preprocessing yielded a lower SC value of 0.56, highlighting the effectiveness of the hybrid approach. Spatial analysis revealed that northern and northwestern China exhibited the highest precipitation variability, particularly in the W3 (8-month) and V5 (trend) components, while southern and southeastern regions demonstrated more stable patterns. The results underscore the value of integrating multiscale temporal analysis with neural-based clustering for robust and interpretable regionalization of precipitation. This framework offers substantial potential for informing water resource management, climate adaptation policies, and infrastructure development under future hydroclimatic uncertainty.
12 August 2024 in Mobile Networks and Applications
Accurate prediction of thermal conductivity in (nano-PEG) composites is essential for accelerating thermal management material design. This study develops a hybrid Random Forest (RF) framework optimized using eight evolutionary algorithms, including (PSO), (GA), (WOA), (GWO), (CSA), (FPA), (FA), and (BA). A dataset of 229 experimental observations was used to model thermal conductivity as a function of temperature, PEG molecular weight, nanoparticle concentration, and nanoparticle form. Among evaluated models, the Bat Algorithm-optimized RF (RF-BA) achieved highest predictive efficiency with R2 = 0.995406, MSE = 0.000196, and AARE = 1.291%, while the PSO-optimized model (RF-PSO) demonstrated the fastest optimization runtime (96.9 s) with competitive accuracy. Correlation and SHAP analyses revealed nanoparticle concentration as the dominant factor governing thermal conductivity (correlation coefficient = 0.75), followed by PEG molecular weight (0.56), temperature (0.33), and nanoparticle form (0.24). The results demonstrate that evolutionarily optimized ensemble learning provides a reliable and computationally efficient strategy for thermophysical property prediction in nano-PEG composites, offering a practical alternative to extensive experimental characterization.