Implement Gaussian Naive Bayes

Question

I'm trying to implement Gaussian Naive Bayes in C# for classification of points. I have implemented first part ( http://www.statsoft.com/textbook/naive-bayes-classifier/ ) probability part, but i don't understand how to implement Gaussian Naive Bayes algorithm normal model. This is my code:

class NaiveBayesClassifier
    {
        private List<Point> listTrainPoints = new List<Point>();
        private int totalPoints = 0;

        public NaiveBayesClassifier(List<Point> listTrainPoints) 
        {
            this.listTrainPoints = listTrainPoints;
            this.totalPoints = this.listTrainPoints.Count;
        }

        private List<Point> vecinityPoints(Point p, double maxDist)
        {
            List<Point> listVecinityPoints = new List<Point>();
            for (int i = 0; i < listTrainPoints.Count; i++)
            {
                if (p.distance(listTrainPoints[i]) <= maxDist)
                {
                    listVecinityPoints.Add(listTrainPoints[i]);
                }
            }
            return listVecinityPoints;
        }

        public double priorProbabilityFor(double currentType)
        {
            double countCurrentType = 0;
            for (int i = 0; i < this.listTrainPoints.Count; i++)
            {
                if (this.listTrainPoints[i].Type == currentType)
                {
                    countCurrentType++;
                }
            }

            return (countCurrentType / this.totalPoints);
        }

        public double likelihoodOfXGiven(double currentType, List<Point> listVecinityPoints)
        {
            double countCurrentType = 0;
            for (int i = 0; i < listVecinityPoints.Count; i++)
            {
                if (listVecinityPoints[i].Type == currentType)
                {
                    countCurrentType++;
                }
            }

            return (countCurrentType / this.totalPoints);
        }

        public double posteriorProbabilityXBeing(double priorProbabilityFor, double likelihoodOfXGiven)
        {
            return (priorProbabilityFor * likelihoodOfXGiven);
        }

        public int allegedClass(Point p, double maxDist)
        {
            int type1 = 1, type2 = 2;

            List<Point> listVecinityPoints = this.vecinityPoints(p, maxDist);

            double priorProbabilityForType1 = this.priorProbabilityFor(type1);
            double priorProbabilityForType2 = this.priorProbabilityFor(type2);

            double likelihoodOfXGivenType1 = likelihoodOfXGiven(type1, listVecinityPoints);
            double likelihoodOfXGivenType2 = likelihoodOfXGiven(type2, listVecinityPoints);

            double posteriorProbabilityXBeingType1 = posteriorProbabilityXBeing(priorProbabilityForType1, likelihoodOfXGivenType1);
            double posteriorProbabilityXBeingType2 = posteriorProbabilityXBeing(priorProbabilityForType2, likelihoodOfXGivenType2);

            if (posteriorProbabilityXBeingType1 > posteriorProbabilityXBeingType2)
                return type1;
            else
                return type2;
        }
    }

In this pdf file (Problem 5) is the description of what i need to do ( http://romanager.ro/s.10-701.hw1.sol.pdf ). My work is to implement Gaussina Naive Bayes and kNN algorithms and compare the result on a set of data. Please teach me where and how to implement Gaussian Naive Bayes algorithm.

Thanks!

for some ideas i don't think that someone want bounty from me ... for this part of algorithm i am completely out. You may consider that my thanks will be your rewards for the solution. I will consider any advice as a solution :D — Urmelinho, Mar 29 '12 at 10:34

score 2 · Accepted Answer · edited May 23 '17 at 12:17

2

good example here:

look at sex predictor at bottom of this page as an example with a real data set

Good explanation here:

naive bayes breakdown on stackoverflow

edited May 23 '17 at 12:17

Community

1
1

answered May 18 '12 at 21:31

negEntropy

319
1
5

Implement Gaussian Naive Bayes

1 Answers1