Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.
Sign up to join this community
I have a large dataset (195 features x 20m samples) that I have trained using XGBoost. I would like to see if a genetic algorithm can beat XGBoost since the data has so much noise it is prone to overfitting.
I would like to use a tree-based model so I don't have to standardize the data, and the features do have some interrelationships.
Are there any python packages that have this all done? Ie.that can create trees through a genetic optimization process?
Very new release:
The main objective of the package is to allow creating decision trees that are better in some aspects than trees made by greedy algorithms. The creation of trees is made by genetic algorithm. In order to achive as fast as possible evolution of trees the most time consuming components are wrtitten in Cython. Also there are implemented mechanisms for using old trees to create new ones without need to classify all observations from beggining (currently in developmnet). There is planned to allow multithreading evolution. The created trees should have smaller sizes with comparable accuracy to the trees made by greedy algorithms.
Also worth to be checked:
PyGAD is an open-source Python library for building the genetic algorithm and optimizing machine learning algorithms.
References:
https://hal.inria.fr/hal-01405549/document
https://www.hindawi.com/journals/tswj/2014/468324/
https://www.kdnuggets.com/2018/07/genetic-algorithm-implementation-python.html
