Azure Machine Learning - A simplified way to get started with Model and analytics

In layman terminology, Machine learning stands for - Building a Model from example inputs to make data driven prediction. Machine learning logic revolves around four things - Data, Algorithm, Model and Analysis (prediction or classification). Machine learning workflow is all about - asking the right question(problem statement), preparing data, Selecting an algorithm (Regression or Classification) and training the model.
Machine learning solution requires an infrastructure starting from preparing dataset to deployment. Azure services brings complement solution in the form of Analytics services. Analytics is discovery, interpretation and communication of meaningful patterns in data. Azure analytics provide wide range of services and infrastructure to build solution of any machine learning problem in an efficient and effective way. Below diagram depict high level overview of Azure machine learning environment.

High level overview of Azure Machine learning environment 

Azure ML Studio is heart of Azure machine learning. It is an IDE for Machine learning and supports all phase of solution development and deployment. It facilitates drag and drop facility for data set preparation, data processing and filtering, model and algorithm pullover in no time. Along with enriched drag & drop, it also provide programmatic control and allows R and Python script execution with ease.

Machine Learning flow of events VS Azure ML : As stated earlier solution to any ML problem begins with preparing data, algorithm selection, training model and evaluation. Below is comparison of ML workflow with Azure ML modules which facilities which provides solution in no time with just vanilla drag & drop and interconnect workflow.   

Normal ML Workflow paralleled with Azure ML Modules

Azure ML components and modules : Azure ML studio brings all possible toolkit for Data, Algorithm and Model. Azure ML studio comes with default sample dataset for get started very quickly. Along with datastore it provides tools for transformation and conversion.

Azure ML Core components 

Machine learning components equipped with various model, Algorithm and  transformation strategy. It also provides interface to plugin R and Python script and run along side of these drag & drop widgets. For deployment of solution as services it provides WebService module (Development to deployment at one place).

Auto Price prediction Experiment : Here we will use Azure ML widgets and create an experiment and evaluate accuracy of system with test data.
In this experiment we are using Automobile price data from sample data(bundled in ML Studio) followed by select columns and clean missing data. Once data has been pre-processed we select linear regression algorithm which is feed to train model with training data set. Finally we use test data set to test this model. All these widgets dragged from left panel and inter connect each other.

Azure ML Experiment predicting Auto price   

Test data set result:  On successful run of this experiment, we can visualise result in form of co-efficient of determination.

Automobile experiment evaluation result

 Above experiment creation and evaluation of result demonstrates Azure ML studio provides an easy way to get started with various ML problem statement using Widgets and deployment in the form of webServices.