Integrative analysis: Support Vector Machine (SVM) classification

In this example, we will showcase RamanSPy’s integrability by integrating a Support Vector Machine (SVM) machine learning model for the identification of different bacteria species.

To build the model, we will use the scikit-learn Python framework.

The data we will use is the Bacteria data, which is integrated into RamanSPy.

from sklearn import svm
from sklearn.metrics import accuracy_score, confusion_matrix
from sklearn.utils import shuffle
import seaborn as sns

import ramanspy

First, we will use RamanSPy to load the validation and testing bacteria datasets.

dir_ = r"../../../../data/bacteria_data"

X_train, y_train = ramanspy.datasets.bacteria("val", folder=dir_)
X_test, y_test = ramanspy.datasets.bacteria("test", folder=dir_)

To guide the training, it is important to shuffle the training dataset, which is originally ordered by bacteria species.

X_train, y_train = shuffle(X_train.flat.spectral_data, y_train)

Then, we can simply use scikit-learn’s implementation of SVMs.

svc = svm.SVC()  # initialisation

Training the SVM model on the training dataset.

_ = svc.fit(X_train, y_train)

Testing the trained model on the unseen testing dataset.

y_pred = svc.predict(X_test.flat.spectral_data)
print(f"The accuracy of the SVM model is: {accuracy_score(y_pred, y_test)}")

The accuracy of the SVM model is: 0.767

Confusion matrix:

cf_matrix = confusion_matrix(y_test, y_pred)
_ = sns.heatmap(cf_matrix, annot=True)