Welcome to DataSafari’s Documentation!¶
DataSafari simplifies complex data science tasks into straightforward, powerful oneliners. Whether you’re exploring data, evaluating statistical assumptions, transforming datasets, or building predictive models, DataSafari provides all the tools you need in one package.
Quick Start¶
Getting started with DataSafari is straightforward.
Install it using pip in your terminal:
pip install datasafari
Or install it using Poetry:
poetry add datasafari
Import DataSafari in your Python script:
import datasafari as ds
For detailed installation options, including installing from the source, see the Installation Guide.
Hypothesis Testing? One line.¶
from datasafari.predictor import predict_hypothesis
import pandas as pd
import numpy as np
# Create a sample DataFrame
df_hypothesis = pd.DataFrame({
'Group': np.random.choice(['Control', 'Treatment'], size=100),
'Score': np.random.normal(0, 1, 100)
})
# Perform hypothesis testing
results = predict_hypothesis(df_hypothesis, 'Group', 'Score')
How DataSafari Streamlines Hypothesis Testing:
Automatic Test Selection: Depending on the data types,
predict_hypothesis()
automatically selects the appropriate test. It uses Chisquare, Fisher’s exact test or other exact tests for categorical pairs, and Ttests, ANOVA and others for categorical and numerical combinations, adapting based on group counts, sample size and data distribution. Assumption Verification: Essential assumptions for the chosen tests are automatically checked.
Normality: Normality is verified using tests like ShapiroWilk or AndersonDarling, essential for parametric tests.
Variance Homogeneity: Tests such as Levene’s or Bartlett’s are used to confirm equal variances, informing the choice between ANOVA types.
 Comprehensive Output:
Justifications: Provides comprehensive reasoning on all test choices.
Test Statistics: Key quantitative results from the hypothesis test.
Pvalues: Indicators of the statistical significance of the findings.
Conclusions: Clear textual interpretations of whether the results support or reject the hypothesis.
Machine Learning? You guessed it.¶
from datasafari.predictor import predict_ml
import pandas as pd
import numpy as np
# Create another sample DataFrame for ML
df_ml = pd.DataFrame({
'Age': np.random.randint(20, 60, size=100),
'Salary': np.random.normal(50000, 15000, size=100),
'Experience': np.random.randint(1, 20, size=100)
})
x_cols = ['Age', 'Experience'] # Feature columns
y_col = 'Salary' # Target column
# Find the best models for your data
best_models = predict_ml(df_ml, x_cols, y_col)
How DataSafari Simplifies Machine Learning Model Selection:
 Tailored Data Preprocessing: The function automatically processes various types of data (numerical, categorical, text, datetime), preparing them optimally for machine learning.
Numerical data might be scaled or normalized.
Categorical data can be encoded.
Text data might be vectorized using techniques suitable for the analysis.
 Intelligent Model Evaluation: The function evaluates a variety of models using a composite score that synthesizes performance across multiple metrics, taking into account the multidimensional aspects of model performance.
Composite Score Calculation: Scores for each metric are weighted according to specified priorities by the user, with lower weights assigned to nonpriority metrics (e.g. RMSE over MAE). This composite score serves as a holistic measure of model performance, ensuring that the models recommended are not just good in one aspect but are robust across multiple criteria.
 Automated Hyperparameter Tuning: Once the top models are identified based on the composite score, the pipeline employs techniques like grid search, random search, or Bayesian optimization to finetune the models.
Output of Tuned Models: The best configurations for the models are output, along with their performance metrics, allowing users to make informed decisions about which models to deploy based on robust, empirically derived data.
 Customization Options & Sensible Defaults: Users can define custom hyperparameter grids, select specific tuning algorithms, prioritize models, tailor data preprocessing, and prioritize metrics.
Accessibility: Every part of the process is in the hands of the user, but sensible defaults are provided for ultimate simplicity of use, which is the approach for
datasafari
in general.
DataSafari at a Glance¶
DataSafari is organized into several subpackages, each tailored to specific data science tasks.
The logic behind the naming of each subpackage is inspired by the typical data workflow: exploring and understanding your data, transforming and cleaning it, evaluating assumptions and finally making predictions.  George
Explorers¶
Explore and understand your data in depth and quicker than ever before.
Module 
Description 

Explore a DataFrame and gain a birdseye view of summary statistics, NAs, data types and more. 

Explore numerical variables in a DataFrame and gain insights on distribution characteristics, outlier detection using multiple methods (Zscore, IQR, Mahalanobis), normality tests, skewness, kurtosis, correlation analysis, and multicollinearity detection. 

Explore categorical variables within a DataFrame and gain insights on unique values, counts and percentages, and the entropy of variables to quantify data diversity. 
For example, use explore_num()
to gain detailed insights into numerical features.
from datasafari.explorer import explore_num
import pandas as pd
import numpy as np
df_explorer = pd.DataFrame({
'Age': np.random.randint(20, 60, size=100),
'Income': np.random.normal(50000, 15000, size=100)
})
explore_num(df_explorer, ['Age', 'Income'])
Transformers¶
Clean, encode and enhance your data to prepare it for further analysis.
Module 
Description 

Transform numerical variables in a DataFrame through operations like standardization, logtransformation, various scalings, winsorization, and interaction term creation. 

Transforms categorical variables in a DataFrame through a range of encoding options and basic to advanced machine learningbased methods for uniform data cleaning. 
For example, use transform_cat()
with the 'uniform_smart'
method for advanced, MLbased categorical data cleaning.
from datasafari.transformer import transform_cat
import pandas as pd
df_transformer = pd.DataFrame({
'Category': ['low', 'medium', 'Medium', 'High', 'low', 'high']
})
transformed_df, uniform_columns = transform_cat(
df_transformer,
['Category'],
method='uniform_smart'
)
Evaluators¶
Ensure your data meets the required assumptions for analyses.
Module 
Description 

Evaluate normality of numerical data within groups defined by a categorical variable, employing multiple statistical tests, dynamically chosen based on data suitability. 

Evaluate variance homogeneity across groups defined by a categorical variable within a dataset, using several statistical tests, dynamically chosen based on data suitability. 

Evaluate and automatically categorize the data types of DataFrame columns, effectively distinguishing between ambiguous cases based on detailed logical assessments. 

Evaluate the suitability of statistical tests for a given contingency table by analyzing its characteristics and guiding the selection of appropriate tests. 
For example, use evaluate_normality()
to check if data distribution fits normality, running the most appropriate normality tests and utilizing a consensus mechanism making for a robust decision on normality.
from datasafari.evaluator import evaluate_normality
import pandas as pd
import numpy as np
df_evaluator = pd.DataFrame({
'Data': np.random.normal(0, 1, size=100)
})
normality = evaluate_normality(df_evaluator, 'Data')
Predictors¶
Streamline model building and hypothesis testing.
Module 
Description 

Conduct the optimal hypothesis test on a DataFrame, tailoring the approach based on the variable types and automating the testing prerequisites and analyses, outputting test results and interpretation. 

Streamline the entire process of data preprocessing, model selection, and tuning, delivering optimal model recommendations based on the data provided. 
For example, use predict_ml()
to preprocess your data, tune models and get the top ML models for your data.
from datasafari.predictor import predict_ml
import pandas as pd
import numpy as np
df_ml_predictor = pd.DataFrame({
'Feature1': np.random.rand(100),
'Feature2': np.random.rand(100),
'Target': np.random.randint(0, 2, size=100)
})
ml_results = predict_ml(
df_ml_predictor,
x_cols=['Feature1', 'Feature2'],
y_col='Target'
)
Contact¶
 Connect with me on:
LinkedIn George Dreemer
Website georgedreemer.com
Thank you very much for taking an interest in DataSafari. ❤️  George