def printGraphFromArrays(names,values,avg,title):
    n = len(names)
    x = np.arange(n)
    width = 0.4

    plt.bar(x, values,width,color='black',align='center')
    plt.xticks(x, names)
    plt.title(title)
    plt.show()

def svmClassification(sample_size):
    test_dataset = openMixedDatasets("pickle/test_dataset.pkl")
    train_dataset = openMixedDatasets("pickle/train_dataset.pkl")
    test_labels = openMixedDatasets("pickle/test_labels.pkl")
    train_labels = openMixedDatasets("pickle/train_labels.pkl")

    test_dataset  = test_dataset[0]
    train_dataset = train_dataset[0]

    clf = svm.SVC(kernel='rbf', gamma=0.0001, C=1000000) #C=100000)

    x_train = train_dataset[:sample_size].reshape(sample_size, 100 * 100 *3)

    x_test = test_dataset.reshape(len(test_dataset), 100 * 100 * 3)

    train_labels = train_labels[0][:sample_size]

    clf.fit(x_train,train_labels)

    diff = compareTwoArrays(test_labels[0],clf.predict(x_test),len(test_labels[0]))
    return diff*100

def mlpClassification(sample_size):
    from sklearn.neural_network import MLPClassifier
    from sklearn.metrics import accuracy_score

    test_dataset = openMixedDatasets("pickle/test_dataset.pkl")
    train_dataset = openMixedDatasets("pickle/train_dataset.pkl")
    test_labels = openMixedDatasets("pickle/test_labels.pkl")
    train_labels = openMixedDatasets("pickle/train_labels.pkl")

    test_dataset = test_dataset[0]
    train_dataset = train_dataset[0]

    x_train = train_dataset[:sample_size].reshape(sample_size, 100 * 100 * 3)
    train_labels = train_labels[0][:sample_size]

    x_test = test_dataset.reshape(len(test_dataset), 100 * 100 * 3)

    clf = MLPClassifier(hidden_layer_sizes=(100,100,100,100), alpha=0.0001,
                        solver='sgd',tol=0.01)
    clf.fit(x_train,train_labels)
    test_pred = clf.predict(x_test)

    diff = compareTwoArrays(test_labels[0],test_pred,len(test_labels[0]))
    return diff * 100
    accuracy_score(test_labels[0], test_pred)

def compareTwoArrays(labels, trained,sampleSize):
    diff = [item for item in labels if item not in trained]
    print("Pocet zle urcenych prvkov je: " + str(len(diff)) + " co je " +str(len(diff)/sampleSize)+"% z " + str(sampleSize)+"prvkov")
    return 1 - (len(diff)/sampleSize)

def classification():

    results = []
    size = [50,100,200,1000,5000]

    for s in size:
        print(s)
        results.append(svmClassification(s))

    printGraphFromArrays(size,results,0,str("SVM rbf klas"))