Preprocessing required data for customer service purpose
Project description
preprocessing_pgp
preprocessing_pgp -- The Preprocessing library for any kind of data -- is a suit of open source Python modules, preprocessing techniques supporting research and development in Machine Learning. preprocessing_pgp requires Python version 3.6, 3.7, 3.8, 3.9, 3.10
Installation
To install the current release:
pip install preprocessing-pgp
To install the release with specific version (e.g. 0.1.3):
pip install preprocessing-pgp==0.2.9
To upgrade package to latest version:
pip install --upgrade preprocessing-pgp
Features
1. Vietnamese Naming Functions
1.1. Preprocessing Names
python
>>> from preprocessing_pgp.name.preprocess import basic_preprocess_name
>>> basic_preprocess_name('Phan Thị Thúy Hằng *$%!@#')
Phan Thị Thúy Hằng
1.2. Enrich Vietnamese Names (Pending...)
python
>>> import pandas as pd
>>> from preprocessing_pgp.name.enrich_name import process_enrich
>>> data = pd.read_parquet('/path/to/data.parquet')
>>> enrich_data = process_enrich(data, name_col='name')
Cleansing Takes 0m0s
Enrich names takes 5m10s
>>> enrich_data.columns
Index(['name', 'predict', 'final'], dtype='object')
1.3. Extract customer type from name (New Feature)
In big data platform, user might enter not just there name into the name field but many others.
This module currently support detection of following type:
- customer : The name of the customer
- company : The name of any company related
- biz : The name of any business related
- edu : The name of any type of education related
- medical : The name of any medical related
python
>>> import pandas as pd
>>> from preprocessing_pgp.name.type.extractor import process_extract_name_type
>>> data = pd.read_parquet('/path/to/data.parquet')
>>> extracted_data = process_extract_name_type(data, name_col='name')
Cleansing names takes 0m0s
Formatting names takes 0m0s
Extracting customer's type takes 0m0s
>>> extracted_data.columns
Index(['username', 'customer_type'], dtype='object')
2. Extracting Vietnamese Phones
python
>>> import pandas as pd
>>> from preprocessing_pgp.phone.extractor import process_convert_phone
>>> data = pd.read_parquet('/path/to/data.parquet')
>>> extracted_data = process_convert_phone(data, phone_col='phone')
Converting phones takes 0m1s
>>> extracted_data.columns
Index(['phone', 'is_phone_valid', 'is_mobi', 'is_new_mobi',
'is_old_mobi', 'is_new_landline', 'is_old_landline',
'phone_convert', 'phone_vendor', 'tail_phone_type'],
dtype='object')
3. Verify Vietnamese Card IDs
python
>>> import pandas as pd
>>> from preprocessing_pgp.card.validation import process_verify_card
>>> data = pd.read_parquet('/path/to/data.parquet')
>>> verified_data = process_verify_card(data, card_col='card_id')
Process cleaning card id...
Verifying card id takes 0m3s
>>> verified_data.columns
Index(['card_id', 'clean_card_id', 'is_valid', 'is_personal_id', 'is_passport',
'is_driver_license'],
dtype='object')
4. Extract Information in Vietnamese Address
All the region codes traced are retrieve from Đơn Vị Hành Chính Việt Nam
Apart from original columns of dataframe, we also generate columns with specific meanings:
- cleaned_<address_col> : The cleaned address retrieve from the raw address column
- level 1 : The raw city extracted from the cleaned address
- best level 1 : The beautified city traced from extracted raw city
- level 1 code : The generated city code
- level 2 : The raw district extracted from the cleaned address
- best level 2 : The beautified district traced from extracted raw district
- level 2 code : The generated district code
- level 3 : The raw ward extracted from the cleaned address
- best level 3 : The beautified ward traced from extracted raw ward
- level 3 code : The generated ward code
- remained address : The remaining address not being extracted
python
>>> import pandas as pd
>>> from preprocessing_pgp.address.extractor import extract_vi_address
>>> data = pd.read_parquet('/path/to/data.parquet')
>>> extracted_data = extract_vi_address(data, address_col='address')
Cleansing takes 0m0s
Extracting takes 0m22s
Code generation takes 0m3s
>>> extracted_data.columns
Index(['address', 'cleaned_address', 'level 1', 'best level 1', 'level 2',
'best level 2', 'level 3', 'best level 3', 'remained address',
'level 1 code', 'level 2 code', 'level 3 code'],
dtype='object')
5. Validate email address
A valid email is consist of:
- Large company email's address (@gmail, @yahoo, @outlook, etc.)
- Common email address (contains at least a alphabet character in email's name)
- Education email (can start with a number)
- Not auto-email
Apart from original columns of dataframe, we also generate columns with specific meanings:
- is_email_valid : indicator of whether the email is valid or not
python
>>> import pandas as pd
>>> from preprocessing_pgp.email.validator import process_validate_email
>>> data = pd.read_parquet('/path/to/data.parquet')
>>> validated_data = process_validate_email(data, email_col='email')
Cleansing email takes 0m0s
Validating email takes 0m22s
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Close
Hashes for preprocessing-pgp-0.2.10.post1.tar.gz
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 | e2dea855d2c84fb2bb97ba7f1968a5e6b591645707b4603c6919095562da89eb |
|
| MD5 | b70378605452242557ba39facec2192a |
|
| BLAKE2b-256 | e61d6ea22cf470b7d11d2ace15fe879b8955e5cbddcc8481a18e2c566c965b33 |
Close
Hashes for preprocessing_pgp-0.2.10.post1-py3-none-any.whl
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 | 4fd1965382ed66876dbea086d411e2f6895fe90ab5deeffa8c2942b38a2ea9e4 |
|
| MD5 | 1ec62cb8a3d2b4786fb946bd8bff6805 |
|
| BLAKE2b-256 | b3fbff2d8be3d7c2c8aec9bcdf9a7131c6e339b5987484edcfeea52e7e802a80 |
