Implementing a neural network?

Question

Machine learning

Hello, help me understand!
I started studying neural networks, after which I was given the task of implementing a predictive model for the occurrence of a violation in production (python). There is input data (7 numbers), as I understand it, this will be a classification task and the back propagation method will be implemented. In building the structure of the neural network, I got 7 input neurons, 5 neurons in the hidden layer and 1 output.
Help me understand if I chose the implementation correctly and I will also supply 1 data array for now in = [1, 0, ...] (7 numbers)
What should I submit to train the network, how will it understand that the number calculated by it is good or bad?

Answer 1

The question is written very incomprehensibly, especially if you really started and really studied (and not just looked through one and a half Internet pages with beautiful drawings).
First, what are seven numbers? Where did they come from? Are these seven different parameters of your "production" or seven values ​​of one parameter for which you are making a forecast?
Secondly, the classification task involves a preliminary stage of learning. On what data are you going to learn it? In what form is this data given to you? What's the output? In other words, is this data marked up? But then it is not clear what kind of "predictions" you are going to do?
Thirdly, the "error back propagation method" is the internal "kitchen" of the neural network. Whom - as you write - he "
Fourth, what kind of "implementation" are we talking about? So far, even the formulation of the problem is out of the question. How it is possible to speak about correct or not implementation? Maybe you don't even need a neural network? Or maybe you need to use which LSTM?
"What should I give to train the network, how will it understand that the number calculated by it is good or bad?" - and this is generally a strange question from a person who, as it were, has already "started studying" neural networks.
As it is written in any good book on machine learning, the solution of any problem does not begin with a tool, but with a clear, complete understanding of the problem. Under it, the most appropriate solution method is selected. And you are trying to do the opposite - without giving a description of the problem, asking with which microscope to hammer this nail.
If consultations are necessary - form specific questions, instead of the general reasonings that it is not clear to you.
P.S. And do not produce more of the same questions: Machine learning?

Answer 2

dmshar I understand you, thanks for the kick)
For 2 weeks of study, I still swim very hard, not understanding the specifics of everything.
If you specify the task, it sounds like this: - "Automated system for predicting violations of labor protection by factory workers"
, experience, average test score for labor protection (%), score of the last test (%), number of violations in the last year, number of violations in the last month.
Please help me figure out which solution is the best!