conjurer 0.0.10

Python library to help you to perform magic on your data analytics project

conjurer

Python library to help you to perform magic on your data analytics project; which helps

• EDA (load & check data)
• Automatic machine learning tuning

For detailed background please refer https://github.com/not-so-fat/conjurer/wiki

Install

pip install conjurer

Usage

You can build prediction pipeline from multiple data sources with following simple code.

from conjurer import (
    eda,
    ml
)

# Load CSVs as pandas.DataFrame
df_dict = {
    name: eda.read_csv("{}_training.csv".format(name)
    for name in ["target", "demand_history", "product", "customer"]
}

# Do feature engineering (not implemented)
feature_training, feature_names = engineer_feature(df_dict)

# Automatic lightgbm tuning 
model = ml.tune_cv("lightgbm", "rg", feature_training, "sales_amount", feature_names, 5)

and produce prediction results.

# Load CSV files for test data set as the same data types as training
loader = eda.DfDictLoader(df_dict)
df_dict_test = loader.load({
    name: "{}_test.csv".format(name)
    for name in ["target", "demand_history", "product", "customer"]
})

# Feature generation for test data set (not implemented)
feature_test = generate_feature(df_dict)

# Get prediction on test data set
model.predict(feature_test)

supported ml algorithms

• LightGBM lightgbm (gbm_autosplit.LGBMClassifier or gbm_autosplit.LGBMRegressor)
• XGBoost xgboost (gbm_autosplit.XGBClassfier or gbm_autosplit.XGBRegressor)
• Random Forest random_forest (sklearn.ensemble.RandomForestClassifier or sklearn.ensemble.RandomForestRegressor)
• Lasso / Logistic Regression linear_model (sklearn.linear_model.Lasso or sklearn.linear_model.LogisticRegression)

This module uses CV by sklearn_cv_pandas.RandomizedSearchCV or sklearn_cv_pandas.GridSearchCV to use pandas.DataFrame for arguments