How to implement decision tree with c# (visual stu

Question

I have a decision tree that i need to turn to a code in C#

The simple way of doing it is using if-else statements but in this solution i will need to create 4-5 nested conditions.

I am looking for a better way to do it and so far i read a little bit about rule engines.

Do you have something else to suggest for an efficient way to develop decision tree with 4-5 nested conditions?

Answer 1

I implemented a simple decision tree as a sample in my book. The code is available online here, so perhaps you could use it as an inspiration. A decision is essentially represented as a class that has references to true branch and false branch and contains a function that does the test:

class DecisionQuery : Decision {
  public Decision Positive { get; set; }
  public Decision Negative { get; set; }
  // Primitive operation to be provided by the user
  public Func<Client, bool> Test { get; set; }

  public override bool Evaluate(Client client) {
    // Test a client using the primitive operation
    bool res = Test(client);
    // Select a branch to follow
    return res ? Positive.Evaluate(client) : Negative.Evaluate(client);
  }
}

Here, Decision is a base class that contains Evaluate method and the source contains one additional derived type that contains a final decision of the tree (yes/no). The type Client is a sample input data that you're analysing using the tree.

To create a decision tree, you can write something like:

var tree = new DecisionQuery {
    Test = (client) => client.Income > 40000,
    Positive = otherTree,
    Negative = someOtherTree
  };

If you just want to write five nested static if clauses then maybe just writing if is fine. The benefit of using a type like this one is that you can easily compose trees - e.g. reuse a part of a tree or modularize the construction.

Answer 2

Below is the Tomas Petricek's code mentioned in the answer https://stackoverflow.com/a/3889544/5288052 .

The zip containing all the source code of the book "Real-World Functional Programming" is available here https://www.manning.com/books/real-world-functional-programming .

// Section 8.4.2 Decision trees in C#

// Listing 8.15 Object oriented decision tree (C#)

abstract class Decision {
  // Tests the given client 
  public abstract void Evaluate(Client client);
}

class DecisionResult : Decision {
  public bool Result { get; set; }
  public override void Evaluate(Client client) {
    // Print the final result
    Console.WriteLine("OFFER A LOAN: {0}", Result ? "YES" : "NO");
  }
}


// Listing 8.16 Simplified implementation of Template method
class DecisionQuery : Decision {
  public string Title { get; set; }
  public Decision Positive { get; set; }
  public Decision Negative { get; set; }
  // Primitive operation to be provided by the user
  public Func<Client, bool> Test { get; set; }

  public override void Evaluate(Client client) {
    // Test a client using the primitive operation
    bool res = Test(client);
    Console.WriteLine("  - {0}? {1}", Title, res ? "yes" : "no");
    // Select a branch to follow
    if (res) Positive.Evaluate(client);
    else Negative.Evaluate(client);
  }
}

static void MainDecisionTrees()
{
  // The tree is constructed from a query
  var tree =
      new DecisionQuery
      {
        Title = "More than $40k",
        // Test is specified using a lambda function
        Test = (client) => client.Income > 40000,
        // Sub-trees can be 'DecisionResult' or 'DecisionQuery'
        Positive = new DecisionResult { Result = true },
        Negative = new DecisionResult { Result = false }
      };

  // Test a client using this tree
  // Create client using object initializer
  var john = new Client {
      Name = "John Doe", Income = 40000, YearsInJob = 1,
      UsesCreditCard = true, CriminalRecord = false 
    };
  tree.Evaluate(john);
}

private static void Main(string[] args)
{
  MainDecisionTrees();
}

Answer 3

Just because... I had a go at this and the results are here ... https://github.com/jkennerley/DeeTree

Answer 4

I had to implement a decision tree using the ID3 algorithm during my masters in C#.

I wrote about my implementation here. The code can be downloaded from GitHub.