rzarno / ml-in-php-start-templates

This project is supposed to be a starting template for replacing business logic in PHP projects with machine learning models. I add here common use cases of how to use machine learning models in PHP projects. In order to use it in your projects you need to replace the data with your dataset and adjust data import and feature engineering stages.

ml-in-php-start-templates

This project is supposed to be a starting template for replacing business logic in PHP projects with machine learning models. I add here common use cases of how to use machine learning models in PHP projects. In order to use it in your projects you need to replace the data with your dataset and adjust data import and feature engineering stages.

Examples cover 4 use cases:

1. Regression model train and use as micorservice

Predict car sale prices based on archive results.

Python + Scikit-Learn + Flask + Gunicorn + Nginx

setup containers:

cd python-docker-regression && docker-compose -f docker-compose.yml up -d --build

request API GET:

http://127.0.0.1:1338/?data=[[22.0, 7.2500, 0, 1, 0, 3, 2.0]]

2. Classification model train and use as micorservice

Classify user agent into indexing bots and others.

Python + Scikit-Learn + Flask + Gunicorn + Nginx

setup containers:

cd python-docker-classify && docker-compose -f docker-compose.yml up -d --build

request API GET:

http://127.0.0.1:1337/?data=Googlebot-Image/1.0

3. Prognose sales from time series

Prognose sales for future dates based on sales from the past.

PHP + Rindow Neural Networks

To train and test model run:

bin/cli prognose-sales-nn-pipeline

or using docker:

docker run --rm rzarno/ml-in-php-start-templates prognose-sales-nn-pipeline

4. Recognize captcha images

Recognize characters on captcha images using convolutional neural networks

PHP + Rindow Neural Networks + Nvidia DAVE-2 CNN model architecture

To train and test model run:

bin/cli captcha-image-classification-cnn-pipeline

or using docker:

docker run --rm rzarno/ml-in-php-start-templates captcha-image-classification-cnn-pipeline

Machine learning process

Machine learning process is a universal sequence of action that you need to perform on your data in order to build prediction model. Each stage of chain can be represented as a separate class, for example: load data, split to train and test set, ... For implementation chain of responsibility pattern was used. Find below diagram showing whole process: