lara-ml-toolkit

Export Eloquent entities to arrays and CSV's for use in basic Classification ML (KNearestNeigbors, Naive Bayes, etc) where a single classification label suffices.

Intended use

Meant for use with

• PHP-ML Dataset classes
• CSV output for any ML library (mostly Python / TensorFlow) that can accept the following CSV format:

sepal_length,sepal_width,petal_length,petal_width
5.1,3.5,1.4,0.2,setosa
4.9,3,1.4,0.2,setosa
4.7,3.2,1.3,0.2,setosa
4.6,3.1,1.5,0.2,setosa

sepal_length	sepal_width	petal_length	petal_width
5.1	3.5	1.4	0.2	setosa
4.9	3	1.4	0.2	setosa
4.7	3.2	1.3	0.2	setosa
4.6	3.1	1.5	0.2	setosa

Note: there is a column without a header, these are the labels/targets.

(Example from the famous Iris dataset)

What this package does NOT do

• We don't separate test set from learning sets, that's your ML framework's job.

Getting Started

1. Require this project in Composer

2. make sure the class(es) that you want to perform classification on implements and conforms to the MlExportable interface and uses the Exportable Trait

3. Use MlExportService to generate and store the exported CSV in your storage folder storage/app/test.csv

$data = MyModel::query()->take(1000)->get();
$csvString = MlExportService::toCsvString($data);
MlExportService::storeCsv('test.csv',  $csvString);

You should now have a CSV file that is ready for use in most ML frameworks

Or if you want to use ArrayDataset objects with PHP-ML:

$data = MyModel::query()->take(1000)->get();
$dataSet = MlExportService::toArrayDataset($data);

---

To-do / goals

• (more) extensive testing
• Implement PHP-ML as part of this package so we can do basic classification easily from inside Laravel.