CODEWITHSUNDEEP
pythonmachine-learningscikit-learnsvm
Viol1997
Viol1997

Reputation: 173

Variability/randomness of Support Vector Machine model scores in Python's scikitlearn

I am testing several ML classification models, in this case Support Vector Machines. I have basic knowledge about the SVM algorithm and how it works.

I am using the built-in breast cancer dataset from scikit learn.

from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.svm import LinearSVC

Using the code below:

cancer = load_breast_cancer()
X_train, X_test, y_train, y_test = train_test_split(cancer.data, cancer.target, 
                                                    stratify=cancer.target, random_state=42)
clf2 = LinearSVC(C=0.01).fit(X_train, y_train)
clf3 = LinearSVC(C=0.1).fit(X_train, y_train)
clf4 = LinearSVC(C=1).fit(X_train, y_train)
clf5 = LinearSVC(C=10).fit(X_train, y_train)
clf6 = LinearSVC(C=100).fit(X_train, y_train)

When printing the scores as in:

print("Model training score with C=0.01:\n{:.3f}".format(clf2.score(X_train, y_train)))
print("Model testing score with C=0.01:\n{:.3f}".format(clf2.score(X_test, y_test)))
print("------------------------------")
print("Model training score with C=0.1:\n{:.3f}".format(clf3.score(X_train, y_train)))
print("Model testing score with C=0.1:\n{:.3f}".format(clf3.score(X_test, y_test)))
print("------------------------------")
print("Model training score with C=1:\n{:.3f}".format(clf4.score(X_train, y_train)))
print("Model testing score with C=1:\n{:.3f}".format(clf4.score(X_test, y_test)))
print("------------------------------")
print("Model training score with C=10:\n{:.3f}".format(clf5.score(X_train, y_train)))
print("Model testing score with C=10:\n{:.3f}".format(clf5.score(X_test, y_test)))
print("------------------------------")
print("Model training score with C=100:\n{:.3f}".format(clf6.score(X_train, y_train)))
print("Model testing score with C=100:\n{:.3f}".format(clf6.score(X_test, y_test)))

When I run this code, I get certain scores per different regularization parameter C. When I would run the .fit lines again (aka train them again), these scores turn out completely different. Sometimes they are even way different (e.g. 72% vs. 90% for the same value of C).

Where does this variability come from? I thought that, assuming I use the same random_state parameter, it would always find the same support vectors and hence would give me the same results, but since the score changes when I train the model another time, this is not the case. In logistic regression, for instance, the scores are always consistent, no matter if I run the fit. code again.

Explaining this variability in accuracy score would be of much help!

Upvotes: 1

Views: 459

Answers (1)

seralouk
seralouk

Reputation: 33147

Of course. You need to fix the random_state=None to a specific seed so that you can reproduce the results.

Otherwise, you use the default random_state=None and thus, every time you call the commands, a random seed is used and this is why you get this variability.

Use:

cancer = load_breast_cancer()
X_train, X_test, y_train, y_test = train_test_split(cancer.data, cancer.target, 
                                                    stratify=cancer.target, random_state=42)
clf2 = LinearSVC(C=0.01,random_state=42).fit(X_train, y_train)
clf3 = LinearSVC(C=0.1, random_state=42).fit(X_train, y_train)
clf4 = LinearSVC(C=1,   random_state=42).fit(X_train, y_train)
clf5 = LinearSVC(C=10,  random_state=42).fit(X_train, y_train)
clf6 = LinearSVC(C=100, random_state=42).fit(X_train, y_train)

Upvotes: 1

Related Questions