Name Classifier Module/Class

This module (class) is a NameClassifier class, and it classifies where a person's name originated from. Naive Bayes algorithm is used and impletented with scikit-learn package. It's still in a development / prototyping phase, and might have some error / bug.

Dependencies

This class utilizes libraries, such as

• scikit-learn
• pandas
• pickle

So make sure all of these packages are installed, as well as thier dependencies.

How to use??

This class can be imported within a python script, or interpreter. There are 2 main use cases,

1. You want to train the model with your own data, from scratch.
2. You want to load pre-trained model, and deploy it.

1. Train your own

1. edit region_list.txt to add or remove the country of your choice, from Faker documentation's locale region codes and country names.
2. generate the data using create_data.py, specifying output and country list file name
3. trian, predict, test and visualize using the module from model.py

2. Pre-trained Model

1. import NameClassifier from model.py
2. use .load_model(fileName.pickle) method to load the model`

Files

• Multi Class Name Classification with Naive Bayes.ipynb
  • goes over how to perform multiclass name classification with NameClassifier class.
• Name Classification with Naive Bayes.ipyn
  • binary classification for Japanese and non-Japanese name
• NameClassifier チュートリアル.ipynb
  • same as above, in Japanese
• model.py
  • the module file for NameClassifier
• prep_data.py
  • practice writing data preprocessing class
• preprocess.py
  • practice data preprocessing
• test.py
  • testing script for module.py

List of methods and attributes

Methods

• __init__https://faker.readthedocs.io/en/master/

  • instantiate the class when training from scratchhttps://faker.readthedocs.io/en/master/.

• load_data()

  • given file names, load the data as pandas Dataframe, add column for label, and split the data into train and test set.
    • params:
      • jp_names(str): file name and full path to csv, containing Japanese names.
      • f_names(str): file name and full path to csv, containing non-Japanese names
    • returns: x_train, x_test, y_train, y_test
      • pandas Series: 2 Series, each containing training & test name data
      • ndarray: 2 ndarray, each containing training and test labels.

• train

  • given training data, vectorize the data and train the Naive Bayes classifier.
  • params:
    • X_train(pandas Series): containing name strings for training.
    • y_train(ndarray): containing labels, 1s and 0s for training.

• predict

  • given names' data, predict whether the names are Japanese or not
  • params:
    • names(list/pandas Series): containing strings of names
  • returns:
    • list: list of 1s and 0s, 1 for Japanese and 0 for non-Japanese.

• evaluate

  • evaluate the model with given test data
  • param:
    • names(list/ndarray): of name strings of people
    • labels(ndarray): of name strings, label
  • returns:
    • dictionary: dictionary of model accuray, precision and recall.

• plot_confusion

  • plots the confusion matrix with provided test data
  • param:
    • yt(ndarray): ground truth labels
    • prediction_test(ndarray): predicted labels, integer

• load_model

  • Load the saved model from pickle file
  • Param:
    • file_name(str): the file name of the model you want to load, including the path to the file,

• save_model(self, file_name)

  • Saves the class using pickle.
  • Params:
    • file_name(str): file name including the path to the file and extension(.pickle)