hyncica / php8-ml
PHP-ML - Machine Learning library for PHP
Requires
- php: ^7.2 || ^8.0
Requires (Dev)
- phpbench/phpbench: ^1.0.0
- phpstan/phpstan: ^0.12
- phpstan/phpstan-phpunit: ^0.12
- phpstan/phpstan-strict-rules: ^0.12
- phpunit/phpunit: ^8.0
- symplify/easy-coding-standard: ^9.4
Replaces
- php-ai/php-ml: v0.9.0
This package is auto-updated.
Last update: 2024-04-27 13:30:57 UTC
README
</a>
This is a port of PHP-ML library (https://gitlab.com/php-ai/php-ml) to make it work on PHP 8. The original library was created by Arkadiusz Kondas (@ArkadiuszKondas).
This port has been created by Michal Hynčica
PORT NOTES
Done during port
- The error when
Phpml\Classification\Ensemble\Bagging
was creating thePhpml\Classification\DecisionTree
instance has been fixed - Fixed expected result in
SetTest::testToArray()
to reflect introduction of stable sorting in php 8. - The required version of
phpbench/phpbench
in composer.json has been changed to^1.0.0
so it can be installed on php 8. The phpbench isn't working currently. Its config need some more work. - The required version of
symplify/easy-coding-standards
has been increased to^9.4
. The config has been converted for newer version.
Failing test
The test DecisionStumpTest::testPredictSingleSample
is currently failing.
This is caused by introduction of stable sorting
in php 8. The problem is in Phpml\Classification\Classifier\OneVsRest::predictSample
method. For non-binary decision stump this method gets probability
for all classifiers. It uses arsort()
to order label by its probability then
return the first label as most probable result.
In test case the probabilities looks like this:
[
'0' => 0.0,
'1' => 1.0,
'2' => 1.0,
];
Before introduction of stable sorting the order of elements with same value wasn't well defined and by some luck the reverse sorted array looked like this:
[
'2' => 1.0,
'1' => 1.0,
'0' => 0.0,
];
But, with stable sorting introduced the reverse sorted array looks like this:
[
'1' => 1.0,
'2' => 1.0,
'0' => 0.0,
];
This is the cause the result is different from what is expected in test.
Installation
Currently this library is in the process of being developed, but You can install it with Composer:
composer require hyncica/php8-ml
The rest is content of original README.md file.
PHP-ML requires PHP >= 7.2.
Simple example of classification:
require_once __DIR__ . '/vendor/autoload.php';
use Phpml\Classification\KNearestNeighbors;
$samples = [[1, 3], [1, 4], [2, 4], [3, 1], [4, 1], [4, 2]];
$labels = ['a', 'a', 'a', 'b', 'b', 'b'];
$classifier = new KNearestNeighbors();
$classifier->train($samples, $labels);
echo $classifier->predict([3, 2]);
// return 'b'
Awards
Documentation
To find out how to use PHP-ML follow Documentation.
Examples
Example scripts are available in a separate repository php-ai/php-ml-examples.
Datasets
Public datasets are available in a separate repository php-ai/php-ml-datasets.
Features
- Association rule learning
- Classification
- SVC
- k-Nearest Neighbors
- Naive Bayes
- Decision Tree (CART)
- Ensemble Algorithms
- Bagging (Bootstrap Aggregating)
- Random Forest
- AdaBoost
- Linear
- Adaline
- Decision Stump
- Perceptron
- LogisticRegression
- Regression
- Clustering
- Metric
- Accuracy
- Confusion Matrix
- Classification Report
- Regression
- Workflow
- Pipeline
- FeatureUnion
- Neural Network
- Cross Validation
- Feature Selection
- Preprocessing
- Normalization
- Imputation missing values
- LabelEncoder
- LambdaTransformer
- NumberConverter
- ColumnFilter
- OneHotEncoder
- Feature Extraction
- Token Count Vectorizer
- NGramTokenizer
- WhitespaceTokenizer
- WordTokenizer
- Tf-idf Transformer
- Token Count Vectorizer
- Dimensionality Reduction
- PCA (Principal Component Analysis)
- Kernel PCA
- LDA (Linear Discriminant Analysis)
- Datasets
- Models management
- Math
Contribute
- Guide: CONTRIBUTING.md
- Issue Tracker: github.com/php-ai/php-ml
- Source Code: github.com/php-ai/php-ml
You can find more about contributing in CONTRIBUTING.md.
License
PHP-ML is released under the MIT Licence. See the bundled LICENSE file for details.
Author
Arkadiusz Kondas (@ArkadiuszKondas)