I am trying to develop a NN or SVM model to predict surface temperature based on several features such as air temp, humidity, wind speed, sunshine etc. I have collected data so I do have the true surface temperature values. I have already tried regression but I also want to try non-parametric models. I am pretty sure that the month, time of day and the day in the month are also factors that can help predict pavement temperature in a more accurate manner. I am using Weka for this purpose. My question is that in creating the NN model can I simply add these 3 new features (time, day, month) as new columns in the training set? Will the model realize that it is dealing with time and date? Any help is greatly appreciated!
Be careful. If you try to predict the temperature from other weather-related predictors (e.g. wind speed, sunshine, humidity, ambient temperature) the information about the period of the year and the day (system state) is already included into the predictors. Unless you think that the relationship between weather variables and surface temperature changes during the year you don't need to include calendar data into your inputs, it might be redundant and then reducing the modelling/prediction performance of your method.