randomForest.py

import numpy
import pandas
import random
from decisionTree import buildDecisionTree, decisionTreePredictions

def trainTestSplit(dataFrame, testSize):
    if isinstance(testSize, float):
        testSize = round(testSize * len(dataFrame))
    indices = dataFrame.index.tolist()
    testIndices = random.sample(population = indices, k = testSize)
    dataFrameTest = dataFrame.loc[testIndices]
    dataFrameTrain = dataFrame.drop(testIndices)
    return dataFrameTrain, dataFrameTest

def bootstrapSample(dataFrame, bootstrapSize):
    randomIndices = numpy.random.randint(low = 0, high = len(dataFrame), size = bootstrapSize)
    return dataFrame.iloc[randomIndices]

def createRandomForest(dataFrame, bootstrapSize, randomAttributes, randomSplits, forestSize = 20, treeMaxDepth = 1000):
    forest = []
    for i in range(forestSize):
        bootstrappedDataFrame = bootstrapSample(dataFrame, bootstrapSize)
        decisionTree = buildDecisionTree(bootstrappedDataFrame, maxDepth = treeMaxDepth, randomAttributes = randomAttributes, randomSplits = randomSplits)
        forest.append(decisionTree)
    return forest

def randomForestPredictions(dataFrame, randomForest):
    predictions = {}
    for i in range(len(randomForest)):
        column = "decision tree " + str(i)
        predictions[column] = decisionTreePredictions(dataFrame, randomForest[i])
    predictions = pandas.DataFrame(predictions)
    return predictions.mode(axis = 1)[0]

def calculateAccuracy(predictedResults, category):
    resultCorrect = predictedResults == category
    return resultCorrect.mean()