https://github.com/victormotogna/irislogisticregression

Iris Dataset Logistic Regression - scikit learn version & from scratch
https://github.com/victormotogna/irislogisticregression

data-science iris-dataset logistic-regression python scikit-learn

Last synced: about 2 months ago
JSON representation

Iris Dataset Logistic Regression - scikit learn version & from scratch

• Host: GitHub
• URL: https://github.com/victormotogna/irislogisticregression
• Owner: VictorMotogna
• Created: 2018-12-09T11:47:43.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2019-01-06T13:12:51.000Z (over 7 years ago)
• Last Synced: 2025-02-04T14:21:58.252Z (over 1 year ago)
• Topics: data-science, iris-dataset, logistic-regression, python, scikit-learn
• Language: Python
• Homepage:
• Size: 7.66 MB
• Stars: 3
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Iris Dataset Logistic Regression

## Dataset

  - Iris Dataset - https://archive.ics.uci.edu/ml/datasets/iris

  - 3 classes x 50 instances each

  - labeled by: sepal length, sepal width, petal length, petal width

  

  ![alt text](https://i.imgur.com/oRACM5G.png "Iris Dataset - sepal length/width")

  

### Data set usages in this implementation

  - using all 4 characteristics: 

    - sepal length, sepal width, petal length, petal width

    - `X = iris.data[:, :4]`

  - using only the first characteristics

    - sepal length, sepal width

    - `X = iris.data[:, :2]`

  - using only `setosa` & `virginica`

    - for __Logistic Regression__ we need to consider binary results

    - `setosa = 0`, `virginica = 1`

![alt text](https://i.imgur.com/4kqr17x.png "Logistic Regression Dataset - sepal length/width")

## Implementations

  1. Logistic Regression from scratch

      - use a `sigmoid` function to output a result between 0 & 1

        - `return 1 / (1 + np.exp(-z))`

      - use a loss function with parameters (weights - theta) to compute the best value for them

        - initially pick random values

        - `return (-y * np.log(h) - (1 - y) * np.log(1 - h)).mean()`

      - gradient descent

        - `gradient = np.dot(X.T, (h - y)) / y.shape[0]`

      - predictions

        - `def predict_probs(X, theta): return sigmoid(np.dot(X, theta))`

  2. Logistic Regression from scikit-learn

      - docs: https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html

      - __much faster__ than the scratch implementation

## Necessary tools

  - `python 3.x`

  - `pip3`

  - `matplotlib (pyplot)`

  - `sklearn`

## References

  - https://medium.com/@martinpella/logistic-regression-from-scratch-in-python-124c5636b8ac

  - https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html