Question

PROGRAMMING : Redirect the console output of the code below via PyQT GUI and have the graph generated to be displayed in a PyQt window.

I saw this heart disease detecting program in GitHub, and I was wondering if I could display the generated graph to a GUI using PyQt. I tried displaying it on a PyQt window, and so far it does show up in the pop up window, but instead the graph appears at the python IDE's console. Here's the original code I was playing around with:

#This code performs the classification of heart disease by labeling the predicted values
# in various classes, namely 0 for absence and 1 to 4 for presence and also try
# to check the model performance by comparing it against other Classifiers

from numpy import genfromtxt
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from sklearn.svm import LinearSVC
from sklearn.decomposition import PCA
import pylab as pl
from itertools import cycle
from sklearn import cross_validation
from sklearn.svm import SVC

#Loading and pruning the data
dataset = genfromtxt('cleveland_data.csv',dtype = float, delimiter=',')
#print dataset
X = dataset[:,0:12] #Feature Set
y = dataset[:,13]   #Label Set

#Method to plot the graph for reduced Dimesions
def plot_2D(data, target, target_names):
     colors = cycle('rgbcmykw')
     target_ids = range(len(target_names))
     plt.figure()
     for i, c, label in zip(target_ids, colors, target_names):
         plt.scatter(data[target == i, 0], data[target == i, 1],
                    c=c, label=label)
     plt.legend()
     plt.savefig('Reduced_PCA_Graph')

# Classifying the data using a Linear SVM and predicting the probability of disease belonging to a particular class
modelSVM = LinearSVC(C=0.001)
pca = PCA(n_components=5, whiten=True).fit(X)
X_new = pca.transform(X)

# calling plot_2D
target_names = ['0','1','2','3','4']
plot_2D(X_new, y, target_names)

#Applying cross validation on the training and test set for validating our Linear SVM Model
X_train, X_test, y_train, y_test = cross_validation.train_test_split(X_new, y, test_size=0.4, train_size=0.6, random_state=0)
modelSVM = modelSVM.fit(X_train, y_train)
print("Testing Linear SVC values using Split")
print(modelSVM.score(X_test, y_test))

# printing the Likelihood of disease belonging to a particular class
# predicting the outcome
count0 = 0
count1 = 0
count2 = 0
count3 = 0
count4 = 0
for i in modelSVM.predict(X_new):
        if i == 0:
                count0 = count0+1;
        elif i == 1:
                count1 = count1+1;
        elif i == 2:
                count2 = count2+1;
        elif i == 3:
                count3 = count3+1;
        elif modelSVM.predict(i) ==4:
                count4 = count4+1
total = count0+count1+count2+count3+count4
#Predicting the Likelihood

#Applying the Principal Component Analysis on the data features
modelSVM2 = SVC(C=0.001,kernel='rbf')

#Applying cross validation on the training and test set for validating our Linear SVM Model
X_train1, X_test1, y_train1, y_test1 = cross_validation.train_test_split(X_new, y, test_size=0.4, train_size=0.6, random_state=0)
modelSVM2 = modelSVM2.fit(X_train1, y_train1)
print("Testing with RBF using split")
print(modelSVM2.score(X_test1, y_test1))

#Using Stratified K Fold
skf = cross_validation.StratifiedKFold(y, n_folds=5)
for train_index, test_index in skf:
   # print("TRAIN:", train_index, "TEST:", test_index)
    X_train3, X_test3 = X[train_index], X[test_index]
    y_train3, y_test3 = y[train_index], y[test_index]
modelSVM3 = SVC(C=0.001,kernel='rbf')
modelSVM3 = modelSVM3.fit(X_train3, y_train3)
print("Testing using stratified with K folds")
print(modelSVM3.score(X_test3, y_test3))

If this program was executed, it will generate a scatter graph on the project's console. Now what I want to happen is, the scatter graph should be in a PyQt window. What's the best way around to do that?

Answer 1

The best way was to use matplotlib.backends.backend_qt4agg to display matplotlib plot in pyQt.I have just made a small pyQT program having a canvas widget where the graph will be displayed and a button to run the code.(press the button and you will see graph in few seconds) I put the code u gave in a method called my_model. I changed plot_2D function a bit because now we want to plot on the canvas.I have commented the code so that you can understand it.The print values will be displayed on the terminal.I used pyQt4.

Code:

import sys
from PyQt4 import QtGui
from numpy import genfromtxt
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from sklearn.svm import LinearSVC
from sklearn.decomposition import PCA
import pylab as pl
from itertools import cycle
from sklearn import cross_validation
from sklearn.svm import SVC

from matplotlib.backends.backend_qt4agg import FigureCanvasQTAgg as FigureCanvas
from matplotlib.backends.backend_qt4agg import NavigationToolbar2QT as NavigationToolbar
from matplotlib.figure import Figure

import random

class Window(QtGui.QDialog):
    def __init__(self, parent=None):
        super(Window, self).__init__(parent)

        # a figure instance to plot on
        self.figure = Figure()

        # this is the Canvas Widget that displays the `figure`
        # it takes the `figure` instance as a parameter to __init__
        self.canvas = FigureCanvas(self.figure)

        # this is the Navigation widget
        # it takes the Canvas widget and a parent
        self.toolbar = NavigationToolbar(self.canvas, self)

        # Just some button connected to `my_model` method
        self.button = QtGui.QPushButton('Run my Code')
        self.button.clicked.connect(self.my_model)

        # set the layout
        layout = QtGui.QVBoxLayout()
        layout.addWidget(self.toolbar)
        layout.addWidget(self.canvas)
        layout.addWidget(self.button)
        self.setLayout(layout)

    #Method to plot the graph for reduced Dimesions
    def plot_2D(self,data, target, target_names,ax):
         colors = cycle('rgbcmykw')
         target_ids = range(len(target_names))
         for i, c, label in zip(target_ids, colors, target_names):
             ax.scatter(data[target == i, 0], data[target == i, 1],
                        c=c, label=label)
       
       

    def my_model(self):
        ''' plot some random stuff '''
        #Loading and pruning the data
        dataset = genfromtxt('cleveland_data.csv',dtype = float, delimiter=',')
        #print dataset
        X = dataset[:,0:12] #Feature Set
        y = dataset[:,13]   #Label Set

        # Classifying the data using a Linear SVM and predicting the probability of disease belonging to a particular class
        modelSVM = LinearSVC(C=0.001)
        pca = PCA(n_components=5, whiten=True).fit(X)
        X_new = pca.transform(X)

        # calling plot_2D
        target_names = ['0','1','2','3','4']

        # create an axis
        ax = self.figure.add_subplot(111)

        # discards the old graph
        ax.clear()
        self.plot_2D(X_new, y, target_names,ax)

        # refresh canvas
        self.canvas.draw()

        #Applying cross validation on the training and test set for validating our Linear SVM Model
        X_train, X_test, y_train, y_test = cross_validation.train_test_split(X_new, y, test_size=0.4, train_size=0.6, random_state=0)
        modelSVM = modelSVM.fit(X_train, y_train)
        print("Testing Linear SVC values using Split")
        print(modelSVM.score(X_test, y_test))

        # printing the Likelihood of disease belonging to a particular class
        # predicting the outcome
        count0 = 0
        count1 = 0
        count2 = 0
        count3 = 0
        count4 = 0
        for i in modelSVM.predict(X_new):
                if i == 0:
                        count0 = count0+1;
                elif i == 1:
                        count1 = count1+1;
                elif i == 2:
                        count2 = count2+1;
                elif i == 3:
                        count3 = count3+1;
                elif modelSVM.predict(i) ==4:
                        count4 = count4+1
        total = count0+count1+count2+count3+count4
        #Predicting the Likelihood

        #Applying the Principal Component Analysis on the data features
        modelSVM2 = SVC(C=0.001,kernel='rbf')

        #Applying cross validation on the training and test set for validating our Linear SVM Model
        X_train1, X_test1, y_train1, y_test1 = cross_validation.train_test_split(X_new, y, test_size=0.4, train_size=0.6, random_state=0)
        modelSVM2 = modelSVM2.fit(X_train1, y_train1)
        print("Testing with RBF using split")
        print(modelSVM2.score(X_test1, y_test1))

        #Using Stratified K Fold
        skf = cross_validation.StratifiedKFold(y, n_folds=5)
        for train_index, test_index in skf:
           # print("TRAIN:", train_index, "TEST:", test_index)
            X_train3, X_test3 = X[train_index], X[test_index]
            y_train3, y_test3 = y[train_index], y[test_index]
        modelSVM3 = SVC(C=0.001,kernel='rbf')
        modelSVM3 = modelSVM3.fit(X_train3, y_train3)
        print("Testing using stratified with K folds")
        print(modelSVM3.score(X_test3, y_test3))

if __name__ == '__main__':
    app = QtGui.QApplication(sys.argv)

    main = Window()
    main.show()

    sys.exit(app.exec_())

-----------------------------------------------------

PS: I referred this https://stackoverflow.com/questions/12459811/how-to-embed-matplotlib-in-pyqt-for-dummies for coming up with the solution.

Hope you understood how to display a plot in PyQt window.