shash 0.2.0

Custom data science utilities for model evaluation and data preparation

Shash Package

A custom Python package for data preparation, exploration, splitting, saving/loading datasets, and model evaluation (classification & regression).

---

✨ Features

🔹 Data Preparation & EDA (dataprep.py)

• datacheck(df) Checks for missing/null values, unique counts, and duplicate rows in a DataFrame.
• dataeda(df) Prints dataset overview: head, shape, info, numerical & categorical statistics.
• auto_convert_dates(df) Automatically converts date-like object/string columns to datetime.

---

🔹 Dataset Splitting & Storage (modelprep.py)

• split_sets(features, target, test_val_ratio=0.3, stratify=False) Splits into Train, Validation, and Test sets (with optional stratification).
• save_sets_csv(...) Saves splits into CSV files (../data/processed/ by default).
• load_sets_csv(...) Loads Train/Val/Test sets from CSV files.

---

🔹 Model Evaluation (evaluation.py)

Classification

• evaluate_classifier(y_true, y_pred_labels, y_pred_proba=None, dataset_name="Dataset") Prints Accuracy, Precision, Recall, F1, ROC AUC (if probs available), classification report, and displays confusion matrix.

Regression

• evaluate_regressor(y_true, y_pred, dataset_name="Dataset") Prints MAE, MSE, RMSE, MAPE, R², and displays residuals & true-vs-predicted plots.

---

🔹 Model Runner Wrappers (model_runner.py)

• fit_eval_classifier(model, X_train, y_train, X_val=None, y_val=None, X_test=None, y_test=None) Fits a classifier and evaluates on Train/Val/Test using evaluate_classifier.
• fit_eval_regressor(model, X_train, y_train, X_val=None, y_val=None, X_test=None, y_test=None) Fits a regressor and evaluates on Train/Val/Test using evaluate_regressor.

---

🚀 Installation

Install from PyPI (after publishing):

pip install shash

Or install locally for development:

pip install -e .

---

📌 Usage Examples

Data Preparation

import pandas as pd
from shash.dataprep import datacheck, dataeda, auto_convert_dates

df = pd.read_csv("data/raw/sample.csv")

# Quick checks
print(datacheck(df))
dataeda(df)

# Convert string dates automatically
df = auto_convert_dates(df)

---

Dataset Splitting

from shash.modelprep import split_sets, save_sets_csv, load_sets_csv

X_train, y_train, X_val, y_val, X_test, y_test = split_sets(features, target, stratify=True)
save_sets_csv(X_train, y_train, X_val, y_val, X_test, y_test)

# Later...
X_train, y_train, X_val, y_val, X_test, y_test = load_sets_csv()

---

Model Evaluation

from shash.evaluation import evaluate_classifier, evaluate_regressor
from sklearn.linear_model import LogisticRegression, LinearRegression

# Classification
clf = LogisticRegression()
clf.fit(X_train, y_train)
evaluate_classifier(y_val, clf.predict(X_val), clf.predict_proba(X_val)[:,1], "Validation")

# Regression
reg = LinearRegression()
reg.fit(X_train, y_train)
evaluate_regressor(y_val, reg.predict(X_val), "Validation")

---

Model Runner

from shash.model_runner import fit_eval_classifier, fit_eval_regressor
from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor

clf = RandomForestClassifier()
fit_eval_classifier(clf, X_train, y_train, X_val, y_val, X_test, y_test)

reg = RandomForestRegressor()
fit_eval_regressor(reg, X_train, y_train, X_val, y_val, X_test, y_test)

---

✅ Tests

All tests are written with pytest. Run them with:

poetry run pytest -v

---