axn-utils 0.3.1

A modular set of data science utilities for EDA, cleaning, and more.

🧠 data-science-snippets

data-science-snippets is a modular, production-ready Python library containing curated, reusable utilities used in the day-to-day workflows of senior data scientists and machine learning engineers.

It includes tools for EDA, cleaning, validation, text processing, feature engineering, visualization, model evaluation, time series, and more — organized by task to keep your work clean and efficient.

---

🚀 Features

✅ Covers every major step in the data science lifecycle
✅ Clean, modular structure by task
✅ Built for reusability in real-world projects
✅ Lightweight: only depends on pandas and numpy by default
✅ Compatible with Python 3.9+

---

📁 Folder Structure

data-science-snippets/
├── eda/
│   ├── most_frequent_values.py
│   ├── data_summary.py
│   ├── cardinality_report.py
│   └── basic_statistics.py
├── data_cleaning/
│   ├── missing_data_summary.py
│   ├── outlier_detection.py
│   └── duplicate_removal.py
├── preprocessing/
│   ├── minmax_scaling.py
│   ├── encoding.py
│   └── normalize_columns.py
├── loading/
│   ├── load_csv_with_info.py
│   ├── safe_parquet_loader.py
│   └── load_large_file_chunks.py
├── visualization/
│   ├── missing_data_heatmap.py
│   ├── distribution_plot.py
│   └── correlation_matrix.py
├── feature_engineering/
│   ├── create_datetime_features.py
│   ├── binning.py
│   ├── interaction_terms.py
│   └── rare_label_encoding.py
├── automated_eda/
│   ├── quick_eda_report.py
│   └── profile_report_wrapper.py
├── model_evaluation/
│   ├── classification_report_extended.py
│   ├── confusion_matrix_plot.py
│   ├── cross_validation_metrics.py
│   └── roc_auc_plot.py
├── text_processing/
│   ├── clean_text.py
│   ├── tokenize_text.py
│   └── tfidf_features.py
├── time_series/
│   ├── lag_features.py
│   ├── rolling_statistics.py
│   └── datetime_indexing.py
├── modeling/
│   ├── model_training.py
│   ├── pipeline_builder.py
│   └── hyperparameter_tuner.py
├── serialization/
│   ├── save_model.py
│   ├── load_model.py
│   └── versioned_saver.py
├── data_validation/
│   ├── schema_check.py
│   ├── unique_constraints.py
│   └── value_range_check.py
├── utils/
│   ├── memory_optimization.py
│   ├── execution_timer.py
│   └── logging_setup.py
└── README.md

---

🔹 eda/ – Exploratory Data Analysis

• most_frequent_values.py: Shows the most common (modal) value per column, its frequency, and percent from non-null values.
• data_summary.py: Summarizes dtypes, nulls, uniques, and memory usage for quick inspection.
• cardinality_report.py: Reports high-cardinality columns in categorical features.
• basic_statistics.py: Returns mean, median, min, max, std, and other summary statistics.

---

🔹 data_cleaning/

• missing_data_summary.py: Shows missing value count and percentage per column, along with data types.
• outlier_detection.py: Detects outliers using IQR or Z-score methods.
• duplicate_removal.py: Identifies and removes duplicate rows or records.

---

🔹 preprocessing/

• minmax_scaling.py: Scales numeric values to a [0, 1] range.
• encoding.py: Label encoding and one-hot encoding utilities.
• normalize_columns.py: Z-score standardization and column normalization helpers.

---

🔹 loading/

• load_csv_with_info.py: Loads CSVs and prints metadata like shape, dtypes, and missing values.
• safe_parquet_loader.py: Robust parquet file loader with fallback options.
• load_large_file_chunks.py: Loads large files in chunks with progress reporting.

---

🔹 visualization/

• missing_data_heatmap.py: Visualizes missing values with a Seaborn heatmap.
• distribution_plot.py: Plots distributions of numeric variables.
• correlation_matrix.py: Draws a correlation heatmap of numeric features.

---

🔹 feature_engineering/

• create_datetime_features.py: Extracts features like day, month, year, weekday from datetime columns.
• binning.py: Performs binning (equal-width or quantile) on continuous variables.
• interaction_terms.py: Creates interaction features (e.g., feature1 * feature2).
• rare_label_encoding.py: Groups rare categorical labels into 'Other'.

---

🔹 automated_eda/

• quick_eda_report.py: Generates a summary of shape, dtypes, nulls, basic stats.
• profile_report_wrapper.py: Wrapper for pandas-profiling / ydata-profiling report generation.

---

🔹 model_evaluation/

• classification_report_extended.py: Displays precision, recall, F1 with support for multiple averages.
• confusion_matrix_plot.py: Annotated confusion matrix visual.
• cross_validation_metrics.py: Computes metrics across folds and aggregates results.
• roc_auc_plot.py: Plots ROC curve and calculates AUC score.

---

🔹 text_processing/

• clean_text.py: Removes punctuation, stopwords, numbers, and lowercases text.
• tokenize_text.py: Word and sentence tokenizers with NLTK or spaCy support.
• tfidf_features.py: Builds TF-IDF matrix from text columns.

---

🔹 time_series/

• lag_features.py: Generates lagged versions of a column for time-aware modeling.
• rolling_statistics.py: Rolling mean, median, std, and min/max features.
• datetime_indexing.py: Time-based slicing, filtering, and resampling helpers.

---

🔹 modeling/

• model_training.py: Trains scikit-learn models with optional cross-validation and logging.
• pipeline_builder.py: Builds preprocessing + modeling pipelines using Pipeline or ColumnTransformer.
• hyperparameter_tuner.py: Wraps GridSearchCV or RandomizedSearchCV with easy setup and evaluation.

---

🔹 serialization/

• save_model.py: Saves models and preprocessing pipelines using joblib or pickle.
• load_model.py: Loads models with optional safety checks.
• versioned_saver.py: Automatically versions and saves models or artifacts in organized directories.

---

🔹 data_validation/

• schema_check.py: Validates schema based on expected dtypes and column names.
• unique_constraints.py: Ensures unique values for IDs or compound keys.
• value_range_check.py: Checks for valid value ranges in numeric columns.

---

🔹 utils/

• memory_optimization.py: Downcasts numerical columns to save memory.
• execution_timer.py: Times function execution with decorators or context managers.
• logging_setup.py: Sets up consistent logging configuration for larger projects.

---

📦 Installation

pip install data-science-snippets

Or install from source:

git clone https://github.com/vAndrewKarma/data-science-snippets.git
cd data-science-snippets
pip install -e .

---

🛠️ Usage

from data_science_snippets.eda.most_frequent_values import most_frequent_values
from data_science_snippets.preprocessing.minmax_scaling import scale_dataframe
from data_science_snippets.modeling.pipeline_builder import build_pipeline

most_frequent_values(df)
scale_dataframe(df, columns=["age", "income"])
build_pipeline(...)

---

📚 Requirements

• Python ≥ 3.9
• pandas ≥ 1.5.3
• numpy ≥ 1.24.4
• seaborn ≥ 0.12.2
• matplotlib ≥ 3.6.3

---

🔐 Security

Please see our SECURITY.md for vulnerability disclosure guidelines.

---

👥 Authors

• Vataselu Andrei
• Nicola-Diana Sincaru

---

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.

---

🌟 Contributions

We welcome contributions! If you have a reusable function or snippet that you think belongs in a senior data scientist’s toolkit, feel free to open a pull request.