As far as I know this is not at present possible. This is a criticism of neural network models generally. Recently some work has been published on explanation of ANN results. A Google scholar search should lead you to this literature. Best wishes David Booth

As Professor David Eugene Booth has pointed out, you won't get an equation out of SPSS Multilayer Perceptron. Imagine a model which has a lot (potentially thousands) of statements like "if x1 then y1" and "if x1 and x2 then y2", and so on. They're "black boxes".

Recent advancements point to "Explainable Neural Networks", which are not yet fully explainable.

These articles may help.

https://medium.com/@shagunm1210/the-explainable-neural-network-8f95256dcddb

https://towardsdatascience.com/explainable-deep-neural-networks-2f40b89d4d6f

How do you write an equation for a neural network?

There are three steps to perform in any neural network:

Hi Zali,

Actually, a single layer "Perception" neural network is equivalent to a linear regression model, and it is also possible to create an MLP network equivalent to a Logistic regression model. However, multi-layer perception (MLP) neural networks have the calculations "distributed" over the nodes of the network, so in general it is not easy to interpret the weights in relation to regression parameters.

What is the nature of your regression models, and the "predictor" and "outcome" variables ?

iyi haftasonlari

Gordon

In my study, I did the constraint programming/optimization, not the regression model to get the weights of paths. It can be considered reverse engineering. Details are in the link here: https://docplayer.biz.tr/43371006-Dogrusal-olmayan-kisitli-programlama-ile-yapay-sinir-aglarinin-egitilmesi-ozet.html