Symbolic Aggregate approXimation in Vector Space Model (SAX-VSM)ΒΆ

This example shows how the SAX-VSM algorithm transforms a dataset consisting of time series and their corresponding labels into a document-term matrix using tf-idf statistics. Each class is represented as a tfidf vector. For an unlabeled time series, the predicted label is the label of the tfidf vector giving the highest cosine similarity with the tf vector of the unlabeled time series. It is implemented as pyts.classification.SAXVSM.

SAX-VSM, tf-idf vector for each class (training set), Cosine similarity between tf-idf vectors for each class and tf vectors for each sample (test set)
# Author: Johann Faouzi <>
# License: BSD-3-Clause

import numpy as np
import matplotlib.pyplot as plt
from pyts.classification import SAXVSM
from pyts.datasets import load_gunpoint

# Toy dataset
X_train, X_test, y_train, y_test = load_gunpoint(return_X_y=True)

# SAXVSM transformation
saxvsm = SAXVSM(window_size=15, word_size=3, n_bins=2,
                strategy='uniform'), y_train)
tfidf = saxvsm.tfidf_
vocabulary_length = len(saxvsm.vocabulary_)
X_new = saxvsm.decision_function(X_test)

# Visualize the transformation
plt.figure(figsize=(14, 5))
width = 0.4

plt.subplot(121) - width / 2, tfidf[0],
        width=width, label='Class 1') + width / 2, tfidf[1],
        width=width, label='Class 2')
plt.ylim((0, 7))
plt.xlabel("Words", fontsize=14)
plt.ylabel("tf-idf", fontsize=14)
plt.title("tf-idf vector for each class (training set)", fontsize=15)

n_samples_plot = 8 - width / 2, X_new[:n_samples_plot, 0],
        width=width, label='Class 1') + width / 2, X_new[:n_samples_plot, 1],
        width=width, label='Class 2')
plt.xticks(np.arange(n_samples_plot), y_test[:n_samples_plot], fontsize=14)
plt.ylim((0, 1.2))
plt.xlabel("True label", fontsize=14)
plt.ylabel("Cosine similarity", fontsize=14)
plt.title(("Cosine similarity between tf-idf vectors for each class\n"
           "and tf vectors for each sample (test set)"), fontsize=15)

plt.suptitle("SAX-VSM", y=0.95, fontsize=22)

Total running time of the script: ( 0 minutes 1.150 seconds)

Gallery generated by Sphinx-Gallery