I'm new to eureqa, so this might be a silly question, but I wasn't able to find an answer to this question.
I'm try to find a good classification model on spatially resolved data. At each location data are collect following different paths. The dataset is organized as follow
locID, PathID, target, var1, var2,..., varN
1 10 xyz
1 2 1 xyz
2 10 xyz
2 20 xyz
2 3 1 xyz
......
I would like to find a classifier able to predict which path has a target value of 1 at each location given the other variables. The other aspect is that, not all locations have the same number of path due to physical constraint. My idea is to maximize/minimize a functional at each locID searching over the different path, hence something of the form
Hm, I'm not sure if I understand your problem. Are you trying to find the the value of the target variable on its last row, ie at the max pathID? Also, do any of your other variables change over time? If none of your other variables change over time, I would try to collapse each of the locID series into a single row with added features (eg what the max pathID was), and model the target using that.
Let start saying that my variables do not change over time. Given that, for each location (locID) different variables (the same for all the paths) are collected following different physically equivalent paths. The number paths is limited to 8, but not all the path are available at each location. For instance locID 1 has 2 path while locID 2 has 3 path.
The target variable represents which of the available path available for the selected location exhibited failure.
What I want to do is to find a function that is path dependent (i.e. use vars) given each location search among all the available paths and predict which of them will fail.
for instance assuming the data are
locID, PathID, target, var1
1 10 5
1 2 1 25
2 10 -1
2 20 3
2 3 1 10
the path of failure at each location can be represented as