Kotlin & Gradle framework for exploring the task of multi-objective portfolio optimization with evolutionary algorithms.
Evolutionary algorithms have long been a topic of interest for me, and Kotlin has been a language have looking to learn more about through actual project of decent size. This project is the intersection of those to desires, a way for me to learn about evolutionary algorithmic techniques through an applied project - written in Kotlin - studying the real-world task of portfolio optimization.
This framework aims to provide solutions to the task of portfolio optimization, given some collection of assets and an objective function - finding some collection of assets that maximize the objection function. Usually those functions, such as the Sharpe ratio are trying to optimize multiple objectives such as maximizing expecting return and minimizing risk.
For a little more background, evolutionary algorithms are type of optimization technique inspired by biological evolution. Basically, the algorithm attempts to find good solutions to some optimization function by maintaining population of possible solutions and iteratively producing new "generations" through the process of pruning bad solutions that don't fit the objective function, mutating the population, and then repopulating the population through "breeding" between solutions. The exact objective function, solution representation, and other details are usually implementation details usually determined by the context of the problem at hand. % more implementation details? learn more at..? Either way, the hope is that over time, better generations pressured for leading to better solutions.
This framework provides a wide range of "drag-and-drop" evolutionary algorithmic techniques, objection functions, and a various number of other parameters - all that is needed is some collection of assets and some history of their returns. The algorithm will then spit back a solution in the form of some number of assets and their weight allocations (percentage of some sum of money to put towards each asset) that attempts to maximize the selected objective function.
The framework offers the following evolutionary strategies, objective functions, and other tuneable parameters:
-
Asset Mutator
-
Weight Mutator
-
Fitness Strategy
-
Populator Strategy
-
Objective Functions
-
Misc
Enter the --help flag into the main function to get more details and the parameters they each expect.
Portolve was built using Kotlin _ and Gradle ...