Dear Ademola

If you know the minimum/maximum possible value of each input, then by a linear mapping (simple linear equation) you can normalize the data. For de-normalizing, use the same linear equation but you need to inverse the function.

If the min/max possible values of each variable is unavailable, you can use the min/max value from the entire data set, but not include the testing set!

You may define two different one dimensional vector for this purpose (if programming in c++). Good luck

Dear Ademola,

First you need to determine the type of transformation you used with your original data. Extract the transformation function and then back-transform your outputs. For example, if you have used Box Cox transformation and used a square root transformation of your inputs to normalize; use a square function to back transform your values. Often we use software applications to transform the values, in that case, know the nature of the transformation function and find the appropriate back transformation formula. I recommend checking your transformation function and back transformation functions in Excel spreadsheet for cross validating your results. Also, if you provide the details on the software environment, the type of transformation you used, we can be of more help. Good luck.

First, which method has been used to normalize. Then find if your method has inverse method and ready.