Wind speed modeling and prediction plays a critical role in wind related engineering studies. However, since the data have random behavior, it is difficult to apply statistical approaches with apriori and deterministic parameters. On the other hand, wind speed data have an important feature; extreme transitions from a wind state to a far different one are rare. Therefore, behavioral modeling is possible. Although several studies focus on global parametrization of wind data behavior, the literature in time-wise modeling and prediction is relatively small. In this study, a novel approach for wind speed modeling using the Mycielski algorithm is demonstrated. The algorithm accurately predicts the time variations of wind speed data in the sense of forecasting future values of wind data by analyzing the repeatedness in the history of the data. The prediction precision of the procedure is tested using wind speed data obtained from three different locations of Turkey (Kayseri, Izmir and Antalya). Prediction results with high accuracy are obtained and presented. (C) 2009 Elsevier Ltd. All rights reserved.