error occured during fusion with GLCM and invarient moment feature

Here i combined(fused) two different type of features i.e invarient moment and GLCM. When i classify with individual invarient moment or GLCM i got 98.75 % accuracy but when i combine these two features with following program i got 25% accuracy, what is the problem in my program i dont understand.

clear all;

clc;

tic; %% calculating elapsed time for execution

%% load mat files

test = {load('db3.mat'),load('db5.mat')};

train = {load('db4.mat'),load('db6.mat')};

n1 = cellfun(@fieldnames,test,'un',0);

n2 = cellfun(@fieldnames,train,'un',0);

V1 = cellfun(@(x,y)[x.(y)],test,[n1{:}],'un',0);

V2 = cellfun(@(x,y)[x.(y)],train,[n2{:}],'un',0);

P_test = cell2mat(reshape(cat(1,V1{:}),100,[])); %for 100 classes

P_train = cell2mat(reshape(cat(1,V2{:}),200,[])); %for 100 classes

%% labeling class

train_label=load('train_label_100.txt');

test_label=load('test_label_100.txt');

%% Normalisation by Z - Scores

P_train = zscore(P_train,0,2);

P_test =zscore(P_test,0,2);

%% classfication

predictlabel = knnclassify(P_test, P_train, train_label,2,'cosine','nearest');

cp = classperf(test_label,predictlabel);

Conf_Mat = confusionmat(test_label,predictlabel);

disp(Conf_Mat);

%% % Evaluate Performance

[FPR, TPR,Thr, AUC, OPTROCPT] = perfcurve(predictlabel, test_label,1);

figure,

plot(TPR,FPR,'r-','LineWidth',1);

xlabel('False positive rate')

ylabel('True positive rate')

title('ROC Curve for Classification ')

t = table(FPR, TPR, Thr);

fprintf('\n\n Overall accuracy:%f%%\n',cp.CorrectRate*100);

%% calculating elapsed time for execution

toc