nomemory / markovneat

Markov chains in Java

Markov chains in Java.

Installing

Maven:

<dependency>
  <groupId>net.andreinc</groupId>
  <artifactId>markovneat</artifactId>
  <version>1.8</version>
</dependency>

Gradle:

implementation 'net.andreinc:markovneat:1.8'

You can also create a "fat" jar using the shadowJar gradle tasks:

gradle shadowJar

The jar will be generated in /build/libs/markovneat*.jar.

Example 1 - Modelling a simple discrete-time markov chain

A directed graph is used bellow to picture the state transitions for a Markov Chain.

The states represent whether a hypothetical stock market is exhibiting a bull market, bear market, or stagnant market trend during a given week.

(See Market trends).

With the markovneat library this can be modelled using the following code:

 MChain<String> marketMChain = new MChain<>();

// Transitioning from "BULL" to "BULL" has a 90% chance
marketMChain.add(new MState<>("BULL"), "BULL", 0.9);
// Transitioning from "BULL" to "BEAR" has a 7,5% chance
marketMChain.add(new MState<>("BULL"), "BEAR", 0.075);
// Transitioning from "BULL" to "STAGNANT" has a 2,5% chance
marketMChain.add(new MState<>("BULL"), "STAGNANT", 0.025);

marketMChain.add(new MState<>("BEAR"), "BEAR", 0.8);
marketMChain.add(new MState<>("BEAR"), "BULL", 0.15);
marketMChain.add(new MState<>("BEAR"), "STAGNANT", 0.05);

marketMChain.add(new MState<>("STAGNANT"), "STAGNANT", 0.5);
marketMChain.add(new MState<>("STAGNANT"), "BULL", 0.25);
marketMChain.add(new MState<>("STAGNANT"), "BEAR", 0.25);

marketMChain.generate(10000).forEach(System.out::println);

Output:

STAGNANT
BULL
BULL
BULL
BULL
... and so on