#Purpose
Making a model to classify handwritten numbers (0~9) by SVM
#Env
- OS : Mac
- Module : scikit-learn
- Lang : Python
#Get character dataset from scikit-learn
You can get handwritten numbers (images) from scikit-learn.
You can check contents of the dataset.
from sklearn import datasets
# Read handwritten dataset
digits = datasets.load_digits()
# You can check the image as you want with matplotlib
import matplotlib.pyplot as plt
plt.matshow(digits.images[0], cmap="Greys")
plt.show()
Sample (0 and 8)
#From image to array
Since SVM can not directly read the images, therefore we need to convert each the image into array dataset. You can change the data to array by [object].data.
# array of each image
X = digits.data
# label of each image
y = digits.target
# split train and test data
# Train data (even)
X_train, y_train = X[0::2], y[0::2]
# Test data (odd)
X_test, y_test = X[1::2], y[1::2]
# you can check the train data
print(X)
print(y)
Content of the Train data (arrays (image))
numbers in each array mean darkness of square.
black > gray > white
[[ 0. 0. 5. ... 0. 0. 0.]
[ 0. 0. 0. ... 10. 0. 0.]
[ 0. 0. 0. ... 16. 9. 0.]
...
[ 0. 0. 1. ... 6. 0. 0.]
[ 0. 0. 2. ... 12. 0. 0.]
[ 0. 0. 10. ... 12. 1. 0.]]
Content of the train data (label)
[0 1 2 ... 8 9 8]
#Make a model
Make a model with train data.
# select SVM
from sklearn import svm
clf = svm.SVC(gamma=0.001)
# train the model with train data
clf.fit(X_train, y_train)
#Test
Evaluate the created model with test data.
Return accuracy
accuracy = clf.score(X_test, y_test)
print(f"Accuracy:{accuracy}")
Accuracy:0.9866369710467706
Return prediction of each test data
predicted = clf.predict(X_test)
[1 3 9 7 9 1 3 5 7 9 1 3 5 7 9..... 9 4 7 3 1 0 2 8 0...]
You can check detail (precision/recall/F-score)
import sklearn.metrics as metrics
print("classification report")
print(metrics.classification_report(y_test, predicted))
classification report
precision recall f1-score support
0 1.00 0.99 0.99 88
1 0.98 1.00 0.99 89
2 1.00 1.00 1.00 91
3 1.00 0.98 0.99 93
4 0.99 1.00 0.99 88
5 0.98 0.97 0.97 91
6 0.99 1.00 0.99 90
7 0.99 1.00 0.99 91
8 0.97 0.97 0.97 86
9 0.98 0.97 0.97 91
accuracy 0.99 898
macro avg 0.99 0.99 0.99 898
weighted avg 0.99 0.99 0.99 898