Author/Authors :
فريداي، اكپ نويسنده Industrial Physics Department, Ebonyi State University, Abakaliki Friday, Ekpe J. Elom, Ibeh G. , امچي، اگبو نويسنده Industrial Physics Department, Ebonyi State University, Abakaliki Amechi, Agbo G. , لننوكنت، ازه نويسنده Industrial Physics Department, Ebonyi State University, Abakaliki Innocent, Ezeh M.
Abstract :
Validation of the results using error analysis show that the estimation with sunshine hours for Angstrom-Prescott has error values of mean bias error = 0.0438, root mean square error = 0.9860, mean percentage error = -1.0991, while artificial neural network has mean bias error = 0.0115, root mean square error = 0.2838, mean percentage error = 0.3830 respectively. For maximum temperature estimation with Angstrom-Prescott, mean bias error = 0.4367, root mean square error = 1.6052, mean percentage error = -2.81650, and with artificial neural network, mean bias error = -0.1650, root mean square error = 1.1442, mean percentage error = 0.4326 respectively. The estimation of relative humidity for Angstrom-Prescott mean bias error = 0.1078, root mean square error = 1.2987, mean percentage error = -1.0945, and for artificial neural network, mean bias error = -0.0105, root mean square error = 0.1289, mean percentage error = 0.0917, and the estimation with cloudiness, Angstrom-Prescott has mean bias error = 0.2143, root mean square error = 1.7231, mean percentage error = -2.2521, while artificial neural network has mean bias error = 0.0556, root mean square error = 0.3386 and mean percentage error = -0.1017 respectively. The values of correlation coefficient and coefficient of determination of Angstrom-Prescott and artificial neural network were also carried out.
