10MAY
Azure Machine Learning on Automobile price data

It was Monday morning and pleasant day indeed when all of sudden, I was over euthanistic about “Azure Machine Learning”. In spite of having own free work space, I realized that the work space remains under-utilized and I haven’t got even some time lately to even check on the “Azure Tour”. Today, I felt is the just that day when I will explore “Azure” for first time.

The step by step guidance of conducting first experience is available at https://docs.microsoft.com/en-us/azure/machine-learning/studio/create-experiment. But before you try this experiment, I will recommend you to know Machine learning concepts as following any series of steps won’t help if you understand the real problem that you might solve using it. So, I know little and thanks to “Unleash Machine Learning Techniques” by Packt thought me the theory and my practical assignments every day has continuously helped me to how put ML to practice. So, I doing simple experiment that is already mentioned in “Azure” i.e., on Automobile Price Data.

 

“Microsoft Azure Machine Learning Studio” give access tons of real problem already solved and stored as samples and can help any student to understand the complete workflow rather than grasping codes in R or python that can be utilized to build each step.

1.      Add your new experiment and keep it Blank. Choose Data set of your choice. Best part is “Azure” offers abundance of datasets which we may use for your future experiment. So, this is right place for Intermediate, may be not beginner (3 months in Data analysis).

 

 

You get option to visualize Automobile price data available at right click.

Now, once you have visualized the data, you may wish to pre-process data simply by selecting columns of your choice. While I find out that option, I realized that the left panel list all different functions neatly. You will know few of options if you are from R or Python

 

1.     Next step is even easy if you, by chance, has familiarity with SSAS (SQL Server Analytical Services) or any other software similar to SSAS. So, we connect two boxes to complete the present workflow and next press Launch column selector. Here, I am followed what Azure suggest i.e., exclude column name named normalized-losses.

2.      Next step is to clean the dataset and remove any row that has missing data. This is pretty straight forward.

3.      Again, from SSAS, I know that to run the experiment, I need click at Run button and here, the button comes right in front of your eyes.

4.      Split the dataset into Training and Test. We will predict price of automobile by training the model based on 70% of dataset.

5.      Score the model and compute predicted values and compare it with actual values. You can check figures under Visualise.

6.      Evaluate the model to understand whether results from Linear Regression hold significant for this present experiment.

 

Azure ML simplifies data analytics and give you the right answers. So, it isn't about "How" but rather "What".Among the other questions that it can answer is:

a. Regression - You can estimate the demand or predict sale figures. In short, Regression forecast the future by estimating the relation between variables.

b. Anomaly Detection - This is most lookahead scenario where you detect the unusual data points and additionally, predict the associated risk.

c. Clustering - Understand varied constumer behaviour and drive appropriate pricing and promotion strategies.

d. Classification - Determine category to which the new information belongs to . You get answers like which of the two coupons draw more customers 

Machine learning helps to find pattern in current data so that it can seek out the patterns in the future data.


Your Queries/Comments