RandomForest.cs

using DecisionTree;
using Microsoft.Data.Analysis;
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;

namespace DTModel
{
	internal class RandomForest
	{
		private List<Tree> forest;
		private int numTrees;

		public RandomForest(DataFrame trainDf, string targetCol, int numTrees = 100)
		{
			this.numTrees = numTrees;
			this.forest = new List<Tree>();

			for (int i = 0; i < numTrees; i++)
			{
				DataFrame bootstrapSample = GetBootstrapSample(trainDf);
				Tree tree = new Tree(bootstrapSample, targetCol, minSampleLeaf: 1, maxDepth: 20);
				forest.Add(tree);
				Console.WriteLine($"Trained tree {i + 1}/{numTrees}");
			}
		}

		private DataFrame GetBootstrapSample(DataFrame df)
		{
			Random rand = new Random();
			long rowCount = df.Rows.Count;
			List<int> rowIndices = new List<int>();
			for (int i = 0; i < rowCount; i++)
				rowIndices.Add(rand.Next(0, (int)rowCount));
			return df.Sample(rowIndices.Count);
		}
		public DataFrame Pred(DataFrame inputDf)
		{
			DebugLog($"Forest Prediction started | Rows: {inputDf.Rows.Count}");

			var finalPredictions = new List<string>();
			foreach (var row in inputDf.Rows)
			{
				var votes = forest.Select(tree => tree.PredictRow(row)?.ToString()).ToList();
				var winner = votes.GroupBy(v => v).OrderByDescending(g => g.Count()).First().Key;
				finalPredictions.Add(winner);
			}
			DataFrame output = inputDf.Clone();
			var predCol = new StringDataFrameColumn("Prediction", finalPredictions);
			output.Columns.Add(predCol);

			DebugLog("Forest Prediction completed");

			return output;
		}

		private void DebugLog(string msg)
		{
			Console.WriteLine($"[Forest DEBUG] {msg}");
		}
	}
}