Machine learning

[yahiaoui_asma]

I have a regression supervised learning which I trained with boosting algorithm, it gives a good results, However , I need to predict a formula that takes the features into account
So, What are the types of machine learning algorithms that we can predict a mathematical equation with them?

[Mubsi]

Hi @yahiaoui_asma,

I remember reading on something similar, courtesy The Batch.

You can take at these two links:

• Deep Symbolic Regression for Recurrent Sequences
• Learning Equations for Extrapolation and Control

Cheers,
Mubsi

[yahiaoui_asma]

please how to know if my problem is linear or not
since i have multiple features i couldn’t plot the features in function of target to know the form of my problem

[Christian_Simonis]

Hi there,

I would suggest to start with visualising your data, e.g. in the following way with histograms of features and scatterplots:

• pandas.plotting.scatter_matrix — pandas 1.4.2 documentation

You can directly see whether some features have linear dependencies to your label as well as if there is redundant information in the data (then a PCA might be useful or you manually want to select the most important features).

Also: if you need a formula, you should also think about fitting a model (in simplest case a linear one) and continue with a residuum analysis to check whether the model meets your requirements.

Following sources might be interesting for you:

• model fit possibility with scipy: scipy.optimize.curve_fit — SciPy v1.8.1 Manual
• example for a model fit with some data: time-series-analysis/Stock_TSA.py at master · christiansimonis/time-series-analysis · GitHub

Best regards
Christian