A package for automating machine learning tasks
Project description
Autopilotml
Automated machine learning library for analytics
Installation
pip install autopilotml
Usage
Load data
from autopilotml import load_data, load_database
# For csv files
df = load_data(path = "dataset/titanic_train.csv", csv=True, **kwargs)
# For excel notebook
df = load_data(path = "dataset/titanic_train.xlsx", excel=True, **kwargs)
# To Load data from Database
# This framework supports sqlite, 'mysql', 'postgres', 'MongoDB'
df = load_database(database_type='sqlite', sqlite_db_path = 'database.db', query='select * from employee_table')
Data Preprocessing
from autopilotml import preprocessing
# If changing any values in the dictionary, whole dictionary has to be provided.
df = preprocessing(dataframe=df, label_column='Survived',
missing={
'type':'impute',
'drop_columns': False,
'threshold': 0.25,
'strategy_numerical': 'knn',
'strategy_categorical': 'most_frequent',
'fill_value': None},
outlier={
'method': 'None',
'zscore_threshold': 3,
'iqr_threshold': 1.5,
'Lc': 0.05,
'Uc': 0.95,
'cap': False})
Data Transformation
from autopilotml import transformation
# If the target_transform is true, then the function return 3 objects, (e.g) dataframe, feature encoder and target encoder
# else it will return 2 objects dataframe and feature encoder
df, encoder = transformation(dataframe=df,
label_column='Survived',
type = 'ordinal',
target_transform = False,
cardinality = True,
Cardinality_threshold = 0.3)
Scaling
# Here if target_scaling = True only applicable for regression then it will return 3 objects dataframe, feature scaler and target scaler
from autopilotml import scaling
df, scaler = scaling(df, label_column= 'Survived', type = 'standard', target_scaling = False)
Feature Selecction
from autopilotml import feature_selection
df, selector = feature_selection(dataframe=df, label_column='Survived',
estimator='RandomForestClassifier',
type='rfe', max_features=10,
min_features=2, scoring= 'accuracy',
cv=5)
Model Training
from autopilotml import training
model = training(dataframe=df, label_column='Survived', model_name='SVC', problem_type='Classification',
target_scaler=None, test_split =0.15, hypertune=True, n_epochs=100)
MLFlow - Track the Model Training and model Parameters
!mlflow ui
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
autopilotml-1.0.6.tar.gz
(200.6 kB
view hashes)
Built Distribution
autopilotml-1.0.6-py3-none-any.whl
(207.3 kB
view hashes)
Close
Hashes for autopilotml-1.0.6-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 25c0a45e241ee94c7f02c735ba6198cbb28ebc09cbd3b5a0e44a49bc371d0f75 |
|
MD5 | cc946988c6af00fa5692a9803ac7bbda |
|
BLAKE2b-256 | c3445e451507ba9d77a7ad876118fba0cb372a2a56aeaa73b2e1409a73dd0356 |