Multivariate time-based forecasting question

[plain python]

Heya, I'm trying to predict the position of an object (X Y and Z) in 3D space, given all of the past positional data. The issue is that X, Y, and Z have direct influences on eachother, and if I only used the past positions of X to predict current X, it may be way off. I would need to use the past positions of Y and Z as well to predict X somewhat accurately - is there something that can do this?

Time series forecasting examples in models I've checked don't account for intermingling of X Y and Z influence. This is for an orbital position predictor and as any offset throws the gravity off, I need some sort of solution that uses past data of all coordinates to predict current ones.

This image is the explanation of why all of the variables influence eachother

Multivariate time-based forecasting question

[pearl adder]

Have you tried multivariate kalman filtering? https://www.kalmanfilter.net/kalmanmulti.html

Found an implementation https://github.com/rlabbe/Kalman-and-Bayesian-Filters-in-Python/blob/master/06-Multivariate-Kalman-Filters.ipynb

[plain python]

The term is new to me, I tried solving the problem by shoving all of my data into a single row of a dataframe

Thanks for the prompt, I'll have a look through the stuff you sent

This is a wee bit advanced, are there any prereqs I should have before attempting to understand it

[pearl adder]

@plain python I guess not much other than Kalman Filter itself, here’s one article https://medium.com/@jaems33/understanding-kalman-filters-with-python-2310e87b8f48