Machine learning to determine the user's weight?

Question

data mining

There is a lot of user activity data that is not decrypted and not signed - Parameters V1-V-30
There is a training sample and a test one, and it is necessary to determine the probability of the user's weight exceeding 90 kg.
The metrics are not signed. What is the algorithm for solving such a problem?

Answer 1

Model selection or Gradient boosting .

Answer 2

Any. Based on this explanation, it is not possible to answer the question. Nowadays people like to use gradient boosting on decision trees. Other things being equal, this method often gives the best result.
To begin with, it is better to try to understand the data itself. It may turn out to remove some of them as unnecessary or to form more informative signs on their basis.

Answer 3

E - sign of the sum
V - parameter
N - index number of the parameter (from 1..30)
A - index from all required chains of parameters (for example, user or session ID)
Vm (middle) - average values ​​of training and/or obviously verified data ( reference)
Vd - delta deviation from the reference.
W - weight (the less, the better)
-----------------
Vm[N]=E(V[N])/N //Average weight of each reference value: average standard
Vd[A][N]=abs(Vm[N]-V[A][N]) //Distance from the standard (for each of the required ones)
W[A]=E(Vd[A][N] ) //Total distance (for each of the desired ones) from the standard
W=min(W[A]) //Minimum distance from all (it is also the maximum weight close to the standard)
A - the desired user with the maximum probability to the desired parameters