Text Classification Using Out-of-Core Learning

Machine LearningMachine LearningBeginner
Practice Now

This tutorial is from open-source community. Access the source code

Introduction

This lab provides an example of how to use scikit-learn for text classification using out-of-core learning. The goal is to learn from data that does not fit into main memory. To achieve this, we make use of an online classifier that supports the partial_fit method, which will be fed with batches of examples. To ensure that the feature space remains the same over time, we leverage a HashingVectorizer that will project each example into the same feature space. This is especially useful in the case of text classification where new features (words) may appear in each batch.

VM Tips

After the VM startup is done, click the top left corner to switch to the Notebook tab to access Jupyter Notebook for practice.

Sometimes, you may need to wait a few seconds for Jupyter Notebook to finish loading. The validation of operations cannot be automated because of limitations in Jupyter Notebook.

If you face issues during learning, feel free to ask Labby. Provide feedback after the session, and we will promptly resolve the problem for you.


Skills Graph

%%%%{init: {'theme':'neutral'}}%%%% flowchart RL sklearn(("`Sklearn`")) -.-> sklearn/CoreModelsandAlgorithmsGroup(["`Core Models and Algorithms`"]) sklearn(("`Sklearn`")) -.-> sklearn/DataPreprocessingandFeatureEngineeringGroup(["`Data Preprocessing and Feature Engineering`"]) sklearn(("`Sklearn`")) -.-> sklearn/UtilitiesandDatasetsGroup(["`Utilities and Datasets`"]) ml(("`Machine Learning`")) -.-> ml/FrameworkandSoftwareGroup(["`Framework and Software`"]) sklearn/CoreModelsandAlgorithmsGroup -.-> sklearn/naive_bayes("`Naive Bayes`") sklearn/DataPreprocessingandFeatureEngineeringGroup -.-> sklearn/feature_extraction("`Feature Extraction`") sklearn/UtilitiesandDatasetsGroup -.-> sklearn/datasets("`Datasets`") sklearn/CoreModelsandAlgorithmsGroup -.-> sklearn/linear_model("`Linear Models`") ml/FrameworkandSoftwareGroup -.-> ml/sklearn("`scikit-learn`") subgraph Lab Skills sklearn/naive_bayes -.-> lab-49235{{"`Text Classification Using Out-of-Core Learning`"}} sklearn/feature_extraction -.-> lab-49235{{"`Text Classification Using Out-of-Core Learning`"}} sklearn/datasets -.-> lab-49235{{"`Text Classification Using Out-of-Core Learning`"}} sklearn/linear_model -.-> lab-49235{{"`Text Classification Using Out-of-Core Learning`"}} ml/sklearn -.-> lab-49235{{"`Text Classification Using Out-of-Core Learning`"}} end

Import libraries and define the parser

import itertools
from pathlib import Path
from hashlib import sha256
import re
import tarfile
import time
import sys

import numpy as np
import matplotlib.pyplot as plt
from matplotlib import rcParams

from html.parser import HTMLParser
from urllib.request import urlretrieve
from sklearn.datasets import get_data_home
from sklearn.feature_extraction.text import HashingVectorizer
from sklearn.linear_model import SGDClassifier
from sklearn.linear_model import PassiveAggressiveClassifier
from sklearn.linear_model import Perceptron
from sklearn.naive_bayes import MultinomialNB


class ReutersParser(HTMLParser):
    """Utility class to parse a SGML file and yield documents one at a time."""

    def __init__(self, encoding="latin-1"):
        HTMLParser.__init__(self)
        self._reset()
        self.encoding = encoding

    def handle_starttag(self, tag, attrs):
        method = "start_" + tag
        getattr(self, method, lambda x: None)(attrs)

    def handle_endtag(self, tag):
        method = "end_" + tag
        getattr(self, method, lambda: None)()

    def _reset(self):
        self.in_title = 0
        self.in_body = 0
        self.in_topics = 0
        self.in_topic_d = 0
        self.title = ""
        self.body = ""
        self.topics = []
        self.topic_d = ""

    def parse(self, fd):
        self.docs = []
        for chunk in fd:
            self.feed(chunk.decode(self.encoding))
            for doc in self.docs:
                yield doc
            self.docs = []
        self.close()

    def handle_data(self, data):
        if self.in_body:
            self.body += data
        elif self.in_title:
            self.title += data
        elif self.in_topic_d:
            self.topic_d += data

    def start_reuters(self, attributes):
        pass

    def end_reuters(self):
        self.body = re.sub(r"\s+", r" ", self.body)
        self.docs.append(
            {"title": self.title, "body": self.body, "topics": self.topics}
        )
        self._reset()

    def start_title(self, attributes):
        self.in_title = 1

    def end_title(self):
        self.in_title = 0

    def start_body(self, attributes):
        self.in_body = 1

    def end_body(self):
        self.in_body = 0

    def start_topics(self, attributes):
        self.in_topics = 1

    def end_topics(self):
        self.in_topics = 0

    def start_d(self, attributes):
        self.in_topic_d = 1

    def end_d(self):
        self.in_topic_d = 0
        self.topics.append(self.topic_d)
        self.topic_d = ""

Define the stream of Reuters documents

def stream_reuters_documents(data_path=None):
    """Iterate over documents of the Reuters dataset.

    The Reuters archive will automatically be downloaded and uncompressed if
    the `data_path` directory does not exist.

    Documents are represented as dictionaries with 'body' (str),
    'title' (str), 'topics' (list(str)) keys.

    """

    DOWNLOAD_URL = (
        "http://archive.ics.uci.edu/ml/machine-learning-databases/"
        "reuters21578-mld/reuters21578.tar.gz"
    )
    ARCHIVE_SHA256 = "3bae43c9b14e387f76a61b6d82bf98a4fb5d3ef99ef7e7075ff2ccbcf59f9d30"
    ARCHIVE_FILENAME = "reuters21578.tar.gz"

    if data_path is None:
        data_path = Path(get_data_home()) / "reuters"
    else:
        data_path = Path(data_path)
    if not data_path.exists():
        """Download the dataset."""
        print("downloading dataset (once and for all) into %s" % data_path)
        data_path.mkdir(parents=True, exist_ok=True)

        def progress(blocknum, bs, size):
            total_sz_mb = "%.2f MB" % (size / 1e6)
            current_sz_mb = "%.2f MB" % ((blocknum * bs) / 1e6)
            if _not_in_sphinx():
                sys.stdout.write("\rdownloaded %s / %s" % (current_sz_mb, total_sz_mb))

        archive_path = data_path / ARCHIVE_FILENAME

        urlretrieve(DOWNLOAD_URL, filename=archive_path, reporthook=progress)
        if _not_in_sphinx():
            sys.stdout.write("\r")

        ## Check that the archive was not tampered:
        assert sha256(archive_path.read_bytes()).hexdigest() == ARCHIVE_SHA256

        print("untarring Reuters dataset...")
        tarfile.open(archive_path, "r:gz").extractall(data_path)
        print("done.")

    parser = ReutersParser()
    for filename in data_path.glob("*.sgm"):
        for doc in parser.parse(open(filename, "rb")):
            yield doc

Set up the vectorizer and hold out a test set

## Create the vectorizer and limit the number of features to a reasonable
## maximum
vectorizer = HashingVectorizer(decode_error="ignore", n_features=2**18, alternate_sign=False)

## Iterator over parsed Reuters SGML files.
data_stream = stream_reuters_documents()

## We learn a binary classification between the "acq" class and all the others.
## "acq" was chosen as it is more or less evenly distributed in the Reuters
## files. For other datasets, one should take care of creating a test set with
## a realistic portion of positive instances.
all_classes = np.array([0, 1])
positive_class = "acq"

## Here are some classifiers that support the `partial_fit` method
partial_fit_classifiers = {
    "SGD": SGDClassifier(max_iter=5),
    "Perceptron": Perceptron(),
    "NB Multinomial": MultinomialNB(alpha=0.01),
    "Passive-Aggressive": PassiveAggressiveClassifier(),
}

## test data statistics
test_stats = {"n_test": 0, "n_test_pos": 0}

## First we hold out a number of examples to estimate accuracy
n_test_documents = 1000
X_test_text, y_test = get_minibatch(data_stream, 1000)
X_test = vectorizer.transform(X_test_text)
test_stats["n_test"] += len(y_test)
test_stats["n_test_pos"] += sum(y_test)
print("Test set is %d documents (%d positive)" % (len(y_test), sum(y_test)))

Define a function to get a minibatch of examples

def get_minibatch(doc_iter, size, pos_class=positive_class):
    """Extract a minibatch of examples, return a tuple X_text, y.

    Note: size is before excluding invalid docs with no topics assigned.

    """
    data = [
        ("{title}\n\n{body}".format(**doc), pos_class in doc["topics"])
        for doc in itertools.islice(doc_iter, size)
        if doc["topics"]
    ]
    if not len(data):
        return np.asarray([], dtype=int), np.asarray([], dtype=int)
    X_text, y = zip(*data)
    return X_text, np.asarray(y, dtype=int)

Define a generator function to iterate over minibatches

def iter_minibatches(doc_iter, minibatch_size):
    """Generator of minibatches."""
    X_text, y = get_minibatch(doc_iter, minibatch_size)
    while len(X_text):
        yield X_text, y
        X_text, y = get_minibatch(doc_iter, minibatch_size)

Iterate over mini-batches of examples and update the classifiers

## We will feed the classifier with mini-batches of 1000 documents; this means
## we have at most 1000 docs in memory at any time.  The smaller the document
## batch, the bigger the relative overhead of the partial fit methods.
minibatch_size = 1000

## Create the data_stream that parses Reuters SGML files and iterates on
## documents as a stream.
minibatch_iterators = iter_minibatches(data_stream, minibatch_size)
total_vect_time = 0.0

## Main loop : iterate on mini-batches of examples
for i, (X_train_text, y_train) in enumerate(minibatch_iterators):
    tick = time.time()
    X_train = vectorizer.transform(X_train_text)
    total_vect_time += time.time() - tick

    for cls_name, cls in partial_fit_classifiers.items():
        tick = time.time()
        ## update estimator with examples in the current mini-batch
        cls.partial_fit(X_train, y_train, classes=all_classes)

        ## accumulate test accuracy stats
        cls_stats[cls_name]["total_fit_time"] += time.time() - tick
        cls_stats[cls_name]["n_train"] += X_train.shape[0]
        cls_stats[cls_name]["n_train_pos"] += sum(y_train)
        tick = time.time()
        cls_stats[cls_name]["accuracy"] = cls.score(X_test, y_test)
        cls_stats[cls_name]["prediction_time"] = time.time() - tick
        acc_history = (cls_stats[cls_name]["accuracy"], cls_stats[cls_name]["n_train"])
        cls_stats[cls_name]["accuracy_history"].append(acc_history)
        run_history = (
            cls_stats[cls_name]["accuracy"],
            total_vect_time + cls_stats[cls_name]["total_fit_time"],
        )
        cls_stats[cls_name]["runtime_history"].append(run_history)

        if i % 3 == 0:
            print(progress(cls_name, cls_stats[cls_name]))
    if i % 3 == 0:
        print("\n")

Plot the results

## Plot accuracy evolution
plt.figure()
for _, stats in sorted(cls_stats.items()):
    ## Plot accuracy evolution with #examples
    accuracy, n_examples = zip(*stats["accuracy_history"])
    plot_accuracy(n_examples, accuracy, "training examples (#)")
    ax = plt.gca()
    ax.set_ylim((0.8, 1))
plt.legend(cls_names, loc="best")

plt.figure()
for _, stats in sorted(cls_stats.items()):
    ## Plot accuracy evolution with runtime
    accuracy, runtime = zip(*stats["runtime_history"])
    plot_accuracy(runtime, accuracy, "runtime (s)")
    ax = plt.gca()
    ax.set_ylim((0.8, 1))
plt.legend(cls_names, loc="best")

## Plot fitting times
plt.figure()
fig = plt.gcf()
cls_runtime = [stats["total_fit_time"] for cls_name, stats in sorted(cls_stats.items())]

cls_runtime.append(total_vect_time)
cls_names.append("Vectorization")
bar_colors = ["b", "g", "r", "c", "m", "y"]

ax = plt.subplot(111)
rectangles = plt.bar(range(len(cls_names)), cls_runtime, width=0.5, color=bar_colors)

ax.set_xticks(np.linspace(0, len(cls_names) - 1, len(cls_names)))
ax.set_xticklabels(cls_names, fontsize=10)
ymax = max(cls_runtime) * 1.2
ax.set_ylim((0, ymax))
ax.set_ylabel("runtime (s)")
ax.set_title("Training Times")


def autolabel(rectangles):
    """attach some text vi autolabel on rectangles."""
    for rect in rectangles:
        height = rect.get_height()
        ax.text(
            rect.get_x() + rect.get_width() / 2.0,
            1.05 * height,
            "%.4f" % height,
            ha="center",
            va="bottom",
        )
        plt.setp(plt.xticks()[1], rotation=30)


autolabel(rectangles)
plt.tight_layout()
plt.show()

## Plot prediction times
plt.figure()
cls_runtime = []
cls_names = list(sorted(cls_stats.keys()))
for cls_name, stats in sorted(cls_stats.items()):
    cls_runtime.append(stats["prediction_time"])
cls_runtime.append(parsing_time)
cls_names.append("Read/Parse\n+Feat.Extr.")
cls_runtime.append(vectorizing_time)
cls_names.append("Hashing\n+Vect.")

ax = plt.subplot(111)
rectangles = plt.bar(range(len(cls_names)), cls_runtime, width=0.5, color=bar_colors)

ax.set_xticks(np.linspace(0, len(cls_names) - 1, len(cls_names)))
ax.set_xticklabels(cls_names, fontsize=8)
plt.setp(plt.xticks()[1], rotation=30)
ymax = max(cls_runtime) * 1.2
ax.set_ylim((0, ymax))
ax.set_ylabel("runtime (s)")
ax.set_title("Prediction Times (%d instances)" % n_test_documents)
autolabel(rectangles)
plt.tight_layout()
plt.show()

Summary

In this lab, we learned how to use scikit-learn for text classification using out-of-core learning. We used an online classifier that supports the partial_fit method, which was fed with batches of examples. We also leveraged a HashingVectorizer to ensure that the feature space remained the same over time. We then held out a test set and iterated over mini-batches of examples to update the classifiers. Finally, we plotted the results to visualize the accuracy evolution and training times.

Other Machine Learning Tutorials you may like