WIND SPEED PREDICTION USING DATA MINING APPROACHES: A CASE STUDY OF GÖKÇEADA, TURKEY

Authors : Fatma Kadriye Düden

Pages : 19-23

View : 43 | Download : 41

Publication Date : 2023-12-31

Article Type : Research Paper

Abstract :In this study, the meteorology data set covering the wind speed, humidity, pressure and temperature data between the years 2014-2021 obtained from the Turkish State Meteorological Service is utilized. With this data set, an estimation is made for the Gokceada district in Canakkale-Turkey, with the WEKA software as pressure and temperature inputs and wind speed output. Gaussian Processes, Linear Regression, Multilayer Perceptron, Simple Linear Regression, SMOreg, Kstar, Decision Table, M5P algorithms in WEKA software are used for estimation. It is made for 7 different groups as temperature-pressure-humidity, temperature-pressure, temperature-humidity, humidity-pressure, temperature, pressure and humidity. According to the results, the best estimation for the temperature-pressure-humidity group is found to be 0.999 for the CC (correlation coefficient) value and 0.2994 for the RMSE (root-mean-square error) with the Kstar algorithm. For the temperature-humidity group, the CC value is 0.9607 and the RMSE value is 0.2777. Estimates from the temperature-pressure and humidity-pressure groups is not give accurate results in comparison to the other groups. The CC and RMSE results are obtained from the humidity and pressure groups are found to be 0.9998 and 0.9985, 0.2679 and 0.0464, respectively.
Keywords : Data mining, Kstar, wind speed, prediction, Gokceada Turkey