Cloud cover is an important meteorological parameter that affects the earth’s ra-diative energy balance and precipitation, playing a major role in the hydrologicalcycle. Long records of cloud cover data are the result of human estimations madeduring the synoptic observation hours at each meteorological station. Here, adoptinga homogenization methodology of the cloud cover time series of the Hellenic NationalWeather Service (HNMS) network and using monthly average cloud cover time series,we analyze 6 meteorological stations in the region of Crete, Greece for a period be-tween 1975 and 2004. To analyze these time series data, we propose the generalizedregression neural network (GRNN), a special type of neural network. By taking ad-vantage of several GRNN properties (i.e., a single design parameter and fast learning)and by incorporating several design strategies, the algorithm investigates effectivelythe possible seasonal and trend patterns found in cloud cover time series data. Fore-casting experiment results implemented under various strategies and time horizonsreveal that the GRNN methods outperforms previously adopted forecasting methodbased on exponential smoothing and ARIMA methodology. Our results reveal thatthe selected GRNN specifications may vary across monitoring stations and forecasting horizons.