Metadata-Version: 2.1
Name: sugeno-classifier
Version: 1.0.0
Summary: Implementation of the Sugeno classifier
Home-page: https://github.com/smeyer198/sugeno-classifier
Author: Sven Meyer
License: MIT
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3.8
Description-Content-Type: text/markdown
Requires-Dist: numpy (>=1.19.0)
Requires-Dist: scikit-learn (>=0.24.1)
Requires-Dist: pulp (>=2.4)

# Sugeno classifier

Implementation of the Sugeno classifier, which was presented in the paper "Machine Learning with the Sugeno Integral: The Case of Binary Classification" [[1]](#1).

# Installation

## Dependencies

The Sugeno classifier requires:
* Python (>=3.8)
* NumPy (>=1.19.0)
* scikit-learn (>=0.24.1)
* PuLP (>=2.4)

## User installation

Use the package manager [pip](https://pip.pypa.io/en/stable/) to install the Sugeno classifier.

```bash
pip install sugeno-classifier
```

# Usage

The implementation is compatible to [scikit-learn](https://scikit-learn.org/stable/) and can be used like other algorithms from this library. In order to use the Sugeno classifier, import the class **SugenoClassifier** from the module **sugeno_classifier** from the package **classifier**. Some examples are shown below:

## First example
Use the contructor and the function **fit** to initialize the Sugeno classifier for a given dataset.

```python
>>> from classifier.sugeno_classifier import SugenoClassifier
>>> X = [[1, 3, 2],
...      [2, 1, 3]]
>>> y = [0, 1]
>>> sc = SugenoClassifier()
>>> sc.fit(X, y)
```
Use the function **predict** to classify samples.

```python
>>> Z = [[3, 2, 1],
...      [1, 2, 3]]
>>> sc.predict(Z)
array([0, 1])
```

## Example with hyperparameters

The Sugeno classifier has two hyperparameter, the maxitivity and the margin, which can be set in the constructor. Both can influence the classification performance. See [[1]](#1) for more information.

```python
>>> from classifier.sugeno_classifier import SugenoClassifier
>>> X = [[1, 3, 2],
...      [2, 1, 3]]
>>> y = [0, 1]
>>> sc = SugenoClassifier(maxitivity=2, margin=0.01)
>>> sc.fit(X, y)
```
Again, the function **predict** can be used to classify samples. Note the different output compared to the first example.

```python
>>> Z = [[3, 2, 1],
...      [1, 2, 3]]
>>> sc.predict(Z)
array([1, 1])
```

## Example with different class labels
The classes do not have to be labeled with 0 and 1, they can be any integer numbers or strings. The label, which is smaller in terms of the relation < or lexicographically ordering, is assigned to negative class and the other one to the positive class.

The first example contains the class labels 2 and 4. Label 2 is assigned to the negative class and label 4 is assigned to the positive class because of 2<4.

```python
>>> from classifier.sugeno_classifier import SugenoClassifier
>>> X = [[1, 3, 2],
...      [2, 1, 3]]
>>> y = [2, 4]
>>> sc = SugenoClassifier()
>>> sc.fit(X, y)
>>> Z = [[3, 2, 1],
...      [1, 2, 3]]
>>> sc.predict(Z)
array([2, 4])
```
The second example contains the class labels 'one' and 'two'. Label 'one' is assigned to the negative class and label 'two' is assigned to the positive class because 'one' comes lexicographically first.
```python
>>> from classifier.sugeno_classifier import SugenoClassifier
>>> X = [[1, 3, 2],
...      [2, 1, 3]]
>>> y = ['one', 'two']
>>> sc = SugenoClassifier()
>>> sc.fit(X, y)
>>> Z = [[3, 2, 1],
...      [1, 2, 3]]
>>> sc.predict(Z)
array(['one', 'two'])
```
# License
[MIT](https://choosealicense.com/licenses/mit/)

# References
[1] Sadegh Abbaszadeh and Eyke Hüllermeier. Machine Learning with the Sugeno Integral: The Case of Binary Classication. 2020.

