Introduction: I would like to introduce you to seven automatic machine learning frameworks, I hope they are valuable.
In recent years, the use rate of machine learning has become higher and higher. Models have brought a series of opportunities for enterprises, and they have also left better ideas for the future. However, the modeling process of machine learning is long and complicated, and people are still seeking to deploy more machine learning models.
When companies need to predict a specific data set, traditional methods need to perform the following operations:
1. Processing data
2. Define technical characteristics
3. Choose a model
4. Optimize hyperparameters
5. Training on parameters
There is no algorithm for all tasks, and data analysts need to select and configure algorithms for each specific task.
In addition, in order to prepare the data, the following steps are required:
1. Determine the column type and semantic content
2. Detect cluster allocation and its ranking
For IT companies, spending money and time is not an advantage. Auto Machine Learning is more effective.
Ranking of automatic learning frameworks
The automatic learning framework can automate all or almost all steps and provide accurate predictions for enterprises. Its biggest advantage is that it can free many business process and data analysts from trivial matters and spend their time on the creative aspects of the project.
Gartner once released a data report, which predicted that in 2020, 40% of big data experts will be replaced by automated machine learning.
To this end, we need to plan ahead and start learning automated machine learning frameworks from now on, and choose the best model and the required parameter configuration.
The following are our selection of seven automated machine learning frameworks, I hope you like them.
ML Box
ML Box is a Python-based database that provides the following functions:
1. Pre-read, read, clean and format data;
2. Select specific functions and detect omissions;
3. Optimize hyperparameters
4. Classification and regression of the most advanced models for prediction
5. Make predictions and model analysis
ML Box is most suitable for running on Linux, while Windows and Mac users may encounter a little difficulty when installing.
ML Box GitHub：https://github.com/AxeldeRomblay/MLBox
ML Box 文档：https://mlbox.readthedocs.io/en/latest/
Auto Sklearn
Auto Sklearn 是一个基于贝叶斯优化、元学习和组合构造的自动机器学习框架，用来查找类似的数据片断。
该软件包含有15种分类算法，还有14个预处理特征，用来定义正确的算法并优化其参数，精度超过98%。
Auto Sklearn特别适合中小型数据集，大型数据集的可扩展性略弱。
Auto Sklearn GitHub：https://github.com/automl/auto-sklearn
Auto Sklearn 文档：https://automl.github.io/auto-sklearn/master/
TPOT
TPOT在2018年8月被GitHub列为最受欢迎的自动机器学习框架。TPOT使用遗传算法来搜索特定任务实现的模型。
TPOT可以同时分析数千个管道，并提供Python的接口。
与 Auto Sklern相比，TPOT提供了自己的回归和分类算法。但是，由于它是一个基于基因编程的架构，每次运行相同的任务，模型都可以提供不同的结果。
TPOT GitHub：https://github.com/EpistasisLab/tpot
TPOT 文档：https://automl.info/tpot/
H2O Auto ML
http://h2o.ai
H2O Auto ML框架是深度学习用户的最佳选择，它可以执行大量需要同时执行多行代码之任务。
H2O使用统计机器学习算法，并有阶梯方式提升机器学习和复杂的学习系统。
H2O GitHub：https://github.com/h2oai
H2O 文档：http://docs.h2o.ai/h2o/latest-stable/h2o-docs/automl.html
Auto Keras
https://autokeras.com
Auto Keras是一款开源的深度学习框架，推动贝叶斯算法优化。此框架可以自动搜索复杂模型的体系结构和超参数
Auto Keras使用神经架构搜索（NAS）算法进行搜索，不需要深度学习工程师参与。
Auto Keras GitHub：https://github.com/keras-team/autokeras
Auto Keras 文档：https://autokeras.com/tutorial/overview/
Google Cloud Auto ML
Google Cloud Auto ML是谷歌云开发的自动机器学习与神经网络框架。它的图形用户界面（GUI)非常易于处理模型，特别适合对机器学习知识掌握有限的开发人员，让人们也能够处理业务所需的模型。
值得一提的是，Google Cloud Auto ML并非开源库，使用时需要付费，它的价值取决于训练模型时所花费的时间以及要预测的图片数据。
Google Cloud Auto ML的学习与开发是免费的。
Goolge Cloud ML文档：https://services.google.com/fb/forms/cloudautomlalphaprogram/
TransmogrifAI
https://transmogrif.ai/
TransmogrifAI是基于Apache Spark框架的Salesforce库，用于Scala语言编写的结构化数据。
TransmogrifAI可以帮助开发者实现深度学习型的准确预测，同时将过程缩短100倍以上。TransmogrifAI棤glks支持处理大规模数据集，亦能够处理Scala上的虚拟机集群。
TransmogrifAI GitHub：https://github.com/salesforce/TransmogrifAI
TransmogriAI 文档：https://docs.transmogrif.ai/
小结
自动化机器学习是企业努力提高性能，更快预测模型的重要工具。
By understanding 7 automatic machine learning frameworks, developers can choose according to business needs and scale of operations, and let it complete their own automatic machine learning tasks.