Machine learning. Which algorithm to choose?

Question

Algorithms

Hello! The essence is that - there is a formula x = y*z. y and z are vectors, x respectively too. The whole process occurs in time from 0 to 60 seconds. The result is x, which must be between 0 and 5.
Example:
tyzx
0 10 3 30
1 8 7 56
2 4 2 8
3 2 2 4
4 1 2 3
5 .. .. 0.00...
and so on. When x falls within the specified limit, the calculation stops.
There are many such tables. Now the problem itself. There will be situations when z will not be known in the formula x = y * z and it must somehow be calculated based on previously calculated examples (from those same tables). Tell me which algorithm is suitable for this or technology / tool.
Thank you.

Answer 1

I'm not special, but you can take the average of past results if the following values ​​do not depend on the previous ones

Answer 2

I'm trying to smoke this https://vk.com/wall-11899736_1754 seems to be explained normally

Answer 3

If I correctly understood the question - how to predict the following from several previous z values, then ordinary regression is enough for this (and judging by the examples, then some very small degree - 2 or 3)

Answer 4

the simplest way to solve problems when you need an approximate calculation of an equation with many unknowns is to bring them to a function of the desired parameters, the essence of which is accuracy, i.e. the smaller its value, the more accurately we guessed the parameters, and then the task converges to multi-criteria search and multidimensional optimization (neural networks, for example, their special case).
further development is the solution of this problem by incomplete enumeration, which means with incomplete information about the solution space, which means an attempt to bypass possible traps, such as local minima.