Network model

SHSguy

Hi, 

 

Sorry if this has been posted before :smileyhappy: 

 

Would like to create a simple network model 

 

Using an input value (randomly chosen between 1 and 5) multiplied by a given weight for each individual input that creates an output value each individual input. Then combines the values of the individual inputs into a total value. 

 

Thank you. 

 

 

sgenzer

hmm interesting - where is that image from?

 

Scott

 

kypexin

Hi @SHSguy

 

Do you need this to be built explicitly in a form of RapidMiner process?

Otherwise this task seems just like simple 1-line Excel function maybe...  

SHSguy

Hi, 

 

Thank you for the reply. 

 

The model I showed is just a simple concept of what I need, the real one would have 9 groups each with an average of 5 subcategories with utilizing a Likert scale (1-5) for a period of no less than 10 years (incremented in monthly inputs). I did model it in Excel but my thesis supervisors requested an alternative utilizing machine learning. 

 

 

SHSguy

Hi Scott, 

 

Made the image using draw.io 

SHSguy

Sorry just noticed I forgot to amend the formula in the picture after making some changes. Please ignore the glaring mistake :smileyhappy:

kypexin

Hi @SHSguy

Your description changes the whole picture... The model you have described does not seem to fit a 'classic' neural network architecture, to my understanding, or at least out-of-the-box solutions within RapidMiner.

Can you please describe your problem in a more general way, what is the data you are working with, what are inputs and outputs, what are you trying to predict? Currently it is not much clear how it would fit a certain machine learning algorithm.

SGolbert

Hi @nicolasvdnest,

 

Do you have a label (known value for the output)? If not, a network model doesn't make any sense. If you clarify this point, I have some ideas of what you could make.

 

Regards,

Sebastian

SHSguy

Hi Sebastian,

 

As all the inputs deal with the same value (e.g. currency) I basically want to create a repeatable model that sums the end values of the various inputs over time. For instance, once a month for a period of ten years. 

 

Cheers,

 

kypexin

Hi @SHSguy

 

In this formulation you don't have a classification task, which means you don't need to predict the output parameter value based on different inputs, but rather you just calculate the output value based on different inpout values thru predefined set of arithmetic operations. So, in this sense, there's still no point to apply any machine learning algorithm.

 

I would also reason like this: if you figured out how to make calculations in Excel, why then you'd need any machine learning here, which is anyway more coplex then just Excel calculations? In other words, if you already obtained your result the simplest way, there's no point to make it more complex.

 

But once again, if you could post full original description of your problem, it would be easier to understand whether or not machine learning algorithms are applicable here at all. 

SHSguy

Hi Vlad, 

 

Thank you for the reply. Yes, I am aware the ML does not come to play in the process (yet). Since this is for my Ph.D. thesis we arent allowed to divulge too much in regards to the data itself and processes due to the fact if someone uses it you lose the ability to publish it. 

 

Fortunately, have found a solution for that aspect :smileyhappy: