Written | Sun Jingtao Source | InfoQ

Machine learning, the frontier of big data, is undoubtedly daunting, as only tech geeks and experts in the field of data science are able to navigate machine learning algorithms and techniques, and it used to be out of reach for most businesses and organizations. But that's changing now, and just as standard APIs have simplified application development, machine learning APIs have lowered the barriers to the field, allowing more and more people and organizations to test the waters of machine learning with the help of APIs offered by companies with deep technical pockets.

Machine learning APIs hide the complexity of creating and deploying machine learning models, allowing developers to focus on data mining and user experience. At the same time, commercializing machine learning as a cloud service is a trend today, with companies such as IBM, Microsoft, Google, Amazon, and BigML offering their own Machine Learning-as-a-Service (MLaaS) for business analysts and developers, and a recent article by Khushbu Shah on KDnuggets describes these 5 companies' machine learning APIs.

IBM Watson

Launched in November 2013, the IBM Watson Developer Cloud offers a complete set of APIs that simplify the process of data preparation and make it easier for developers to run predictive analytics. As a cognitive service, the IBM Watson APIs allow developers to leverage machine learning techniques such as natural language processing, computer vision in order to and predictive capabilities to build smarter products, services, or apps, and by embedding IBM Watson in their apps, developers are also able to better understand how users interact with the app.

IBM Watson is an extended toolset of perceptual capabilities for listening, seeing, speaking, and understanding, with more than 25 APIs covering nearly 50 technologies, the most prominent of which include:

Machine Translation - helps translate text in different language combinations

Message **** vibration - to find out the popularity of a phrase or word among a predetermined group of people

Q&A - to provide direct answers to queries triggered by the main document source

User modeling -predicts people's social characteristics based on given text

Microsoft Azure Machine Learning API

Microsoft Azure Machine Learning is a platform for processing massive amounts of data and building predictive applications that provides features such as Natural Language Processing, Recommendation Engine, Pattern Recognition, Computer Vision, and Predictive Modeling, etc. To cater to the preferences of data scientists, the Microsoft Azure Machine Learning platform also adds support for Python, enabling users to publish Python code snippets directly into the API. with the Microsoft Azure Machine Learning API, data scientists can more easily build predictive models and shorten development cycles, with key features including:

Support for creating custom, configurable R modules that allow data analysts or data scientists to use their own R-language code to perform training or prediction tasks

Support for custom Python scripts, which can be written using SciPy, SciKit-Learn, NumPy, and Pandas data science libraries

Support for training on petabytes of data, and support for Spark and Hadoop big data processing platforms

Google Prediction API

The Google Prediction API is a cloud-based machine learning and pattern-matching tool that can draw data from BigQuery and Google Cloud Storage and is able to handle user scenarios such as sales opportunity analysis, customer sentiment analysis, customer churn analysis, spam detection, document categorization, purchase rate prediction, recommendations and smart routing. Users using Google Prediction API do not need knowledge of artificial intelligence, just some basic programming background.Google Prediction API supports numerous programming languages such as .NET, Go, Google Web Toolkit, JavaScript, Objective C, PHP, Python, Ruby and Apps Script, basically covering the mainstream programming languages.

Amazon Machine Learning API

The Amazon Machine Learning API simplifies the process of implementing predictions by enabling users to implement model building, data cleaning, and statistical analysis without the need for a large number of data experts. While the API has some UI or algorithmic limitations, it is user-friendly and wizard-driven, and it provides developers with a number of visualization tools that make the use of the relevant APIs more intuitive and also clearer.

User scenarios supported by the Amazon Machine Learning API include:

Classifying songs by topic by analyzing signal level characteristics

Identifying a user's activity by analyzing data captured by acceleration sensors on smart devices and signals from gyroscopes, whether it's going up the stairs, going down the stairs, lying down, sitting down, or standing still

Advancing the user's behavior by analyzing the user behavior to predict whether a user will become a paying subscriber

Analyze website activity logs to detect fake users, bots, and spammers in the system

BigML

BigML is a user-friendly, developer-friendly machine learning API motivated by a desire to make predictive analytics simpler and more understandable for users. The BigML API provides 3 important modes: a command line interface, a web interface, and a RESTful API, and the main features it supports include anomaly detection, cluster analysis, SunBurst visualization of decision trees, and text analysis.

With BigML, users are able to understand the relationships between individual attributes and predictive attributes in complex data by creating a descriptive model, create predictive models based on past sample data, and maintain models on the BigML platform and use them remotely.

Link to this article:/news/2015/12/5-best-ml-api-to-use