Internet can be used as one important source of information for machine learning algorithms. Web pages store diverse information about multiple domains. One critical problem is how to categorize this information.

Websites classification is performed by using NLP techniques that helps to generate words frequencies for each category and by calculating categories weights it is possible to predict categories for Websites.

Main dataset for this project could be found: URL categorization dataset file

For url predictions you could use already generated words frequency model that was created at 2020-12-26: frequency_mode/word_frequency_2021.pickle

Otherwise, you could create your own model words frequency model by executing construct_features.py.

For python versions management I would highly advise to use pyenv tool.

1. Execute poetry shell. Poetry installation guide

foo@bar:~$ poetry shell

2. Start FastAPI local server

foo@bar:~$ uvicorn url_predictions.api_main:app --reload

3. There are two ways to get website category predictions:

• Use curls commands to predict url:

curl -X 'POST' \
  'http://localhost:8000/predict/?url=bbc.com' \
  -H 'accept: application/json' \
  -d ''

• Use FastAPI UI:
  1. Go to http://localhost:8000/docs
  2. Expand /predict/ POST endpoint page
  3. Write an url and press execute
  4. You should get a JSON response with the results

Request url: http://localhost:8000/predict/?url=bbc.com Response:

{
  "main_category": "News_and_Media",
  "category_weight": 7123532,
  "sub_category": "Reference",
  "sub_weight": 7038726,
  "response": "All HTML content",
  "tokens":  [
      "bbc",
      "homepage",
      "homepageaccessibility",
      "linksskip",
      "contentaccessibility",
      .
      .
      .
      ]
}

Website Classification Using Machine Learning Approaches.pdf

This project is my Bachelor thesis, so it also has documentation part written with LaTeX. Documentation with original LaTeX files is located in the Documentation folder.

Please note that some concepts of documentation does not exist anymore or there are new things since some changes were applied after documentation was written.