lazybrains 0.0.1

Model Selection Tool

lightml

lightml is a simple model selection tool for linear regression, supporting OLS (Normal Equation), Batch Gradient Descent (BGD), and BGD with L1 regularization (Lasso). It is built on top of pandas and numpy.

Usage

1. Import and Prepare Data

import pandas as pd
from lightml import lightml

df = pd.read_csv("your_data.csv")
lm = lightml(df)

2. Split Data

lm.fit_data(
    random_state=42,
    ratio=0.8,
    training_features=['feature1', 'feature2'],
    target_features=['target']
)

3. Standardize Features (Optional)

lm.Standard_Scale(features=['feature1', 'feature2'])

4. Train Models

OLS (Normal Equation)

score = lm.doML(model="lr", method="ols", get_equation=True)
print("R2 Score:", score)

Batch Gradient Descent

score = lm.doML(model="lr", method="bgd", epochs=200, learning_rate=0.01, get_equation=True)
print("R2 Score:", score)

Lasso Regression (BGD + L1)

score = lm.doML(model="lr", method="bgd", penalty="l1", epochs=200, learning_rate=0.01, lamda_=0.1, get_equation=True)
print("R2 Score:", score)

5. Save/Load Model

lm.save_model(model_obj, filename="model.pkl")
loaded_model = lm.load_model(filename="model.pkl")

---

MIT

---