%Image Features Classification
addpath(genpath('/Users/manalhelal/Dropbox/AAST/undergrad/F2014/CC716'));

clear all;
load images_2;
imagesMat = images_2;

%file containing rows of 52 images, and columns of 57 features

%create the Euclidian distance matrix
D = pdist(imagesMat);
%create the Agglomerative hierarchical cluster tree
Z = linkage(D);
%cluster the images into 6 classes
C = cluster(Z,'maxclust',6);

%create the NB Classifier model using only the first 8 features, because the within-class variance of features 9, 21, 24, 50 in class 1 are not positive, and it should be positive for Naive Bayes classifier to work 

NBModel = fitNaiveBayes(imagesMat(:, 1:8),C);

%using the generated model, predict the class of image 50, using the same
%features

pre = predict(NBModel,imagesMat(50, 1:8));

%return the posterioir probability of the same image to all classes, and
%the log of the 
[post,cpre,logp] = posterior(NBModel,imagesMat(50, 1:8));

%draw the 3d scatter plot of the first three features of the images using
%circle size 10, and colouring based on the assigned classes to visualise
%the distribution.

scatter3(imagesMat(:, 1), imagesMat(:, 2), imagesMat(:, 3), 10, C);


% read the estimates of the class, the mean and the standard deviation of
% the predicted class
estimates = NBModel.Params{pre,1};