My experience of passing the Microsoft Azure Machine Learning Exam (70-774)

I recently passed Exam 70-774 and would like to share my experience in passing the exam. The exam mostly focuses on various features of Azure Machine Learning Studio. I have created about 54 pages of notes which have links to relevant topics in the exam. The links are not organized based on topics and you may see a repetition of concepts.

If you read these notes and the material referenced with the links, I think you should pass without any doubt. The passing score is 700 out of 1000. I scored 827. I strongly recommend following the notes (PDF) file, you should be able to understand all the questions and pass the exam.

Books and other resource which I refereed?

• Microsoft Publication - Exam Ref 70-774 Perform Cloud Data Science with Azure Machine Learning. The book is a good read but from exam point of view, its not very helpful. Microsoft Website is your best source. Most of the reference links in the notes PDF are from Microsoft ML Studio website.
• Udemy course - Microsoft Azure Machine Learning Practice Tests (70-774) - Again not very useful.

Who should take this exam?

This exam is oriented to DBAs, Data Scientist, Data Architects, Data Analysts, Data Developers or professional who want to learn or who want to be certified in Data Analysis.

How much time is required to prepare and pass this exam?

If you have a background in machine learning, than 3-4 weeks should be good, else it may take 7-8 weeks.

About the exam: There are 37-42 multiple choice questions in total.

10-11 questions are where you are given a scenario and you are asked to answer but you cannot change the previous answer. It means the question is same but multiple choices are given in separate pages and you cannot go back and correct the answer in previous page. Example question: You have a dataset that contains millions of large digital photographs. You plan to detect the presence of a object in the photographs. First set of question will give various options to detect the object, for example is using "Boosted Decision Tree Regression" meet the goal of identifying the object? You have to answer Yes or No. Second question will be based on the same scenario but different questionnaire. These are tricky questions and you really need to read them carefully.

Concepts on which multiple choice question are based upon:

1. Practice Azure ML Studio as its free.
2. Understand the advantages of "Azure Machine Learning experiment" can be developed and published by other users in the Microsoft development community". Be aware Microsoft ML wants to highlight its strengths.
3. Read about requirement to handle very high volume of data and up to 50x faster speeds, Refer to Azure HDInsight using Microsoft R Server in attached notes. Read about Azure Hadoop cluster and R Server integration with Azure.
4. You should know when to use the python script in Azure ML studio. You should know when to use python in ML Studio and its limitations. You are not required to write python code.
5. Read about Matchbox reommender, Which Matchbox recommender should you use in a given situation?
6. Know about Mean absolute error (MAE) and Relative Absolute Error (RAE)- Understand what does a low or high value means.
7. Basic concept about language R code - you do not need to write the code but should be able to arrange the answers in order of execution. People with no experience in R should focus on the sequence of execution, if you pay attention you should be able to understand. Just follow my link about R in the attached notes. Its alright not knowing R language, but anyone with any prior knowledge of any programming language should be able to understand.
8. Knowing Principal Component Analysis and anomaly detection is very important from exam point of view.
9. There are various modules in Azure Machine Learning Studio - You should know when to use "Select Columns in Dataset", "Select Column Transform", "Convert to Indicator Values", "Edit Metadata", "Execute Python Script", "Tune Model Hyperparameters", "Normalize Data", "Import Data", "Clip Values" and "Clean Missing Data". These are the building blocks of building machine learning models without writing minimum code.
10. Once the model is build, the advantage of any cloud service provider including Azure is that you can "Setup Web-service" and "configure a Web Service", for exam point of view, you should know the steps required to setup a web service. Observe that when there are errors in the model can you create web service?
11. Understand the basic steps of building a model with the help of ML Studio modules. When data is split, trained, validated etc. There are many built in models in the "experiment" section of ML Studio. Run these models yourself and visualize the intermediate data in each step. Observe the input and output of each module while building the model, see errors when required input is missing. This is core to the ML Studio and it is expected that you have solid understanding of the steps while building the model.
12. Any Machine Learning models depends on large amount of data. Think data as the input or food for the model you are feeding in, thus knowing about Apache Spark cluster in Azure HDinsight, Hive - how to prepare data is important to your learning goals as well.
13. SQL Server has integration with R, You should understand how you can use R code in a Transact-SQL statement.
14. There are multiple R packages which is out of the scope of this exam but you should know how to use in SQL Server VM and R packages.
15. In the attached notes, there are various charts which explains which model should you select based on a given situation. For example when to use Multiclass Logistic Regression, Boosted Decision Tree Regression, Decision Forest Regression, Poisson Regression etc. Understand and try to make the chart yourself.
16. Cross-validation is important to know for any machine learning exam or while evaluating the accuracy of two or more models.
17. Understand model over-fitting and Under-fitting.

All the above concepts are very important for any Machine Learning exam. Please refer all the links and notes carefully, practice ML Studio experiments. Try to recreate these models yourself, play around with ML Studio and you should be good to go.

Good Luck and add your comments if the above mentioned points are helpful.