I started Microsoft Azure ML. It is definitely amazing!

 

binary-208216_1280

 

Finally, I started MS (Microsoft) Azure ML (Machine Learning).  So in this blog,  I would like to report what it is and why it is amazing for not only data scientists but also businessmen/women. MS Azure ML is a kind of ML services on the cloud. It is easy to start data analysis by ML, even for beginners.  For data analysis, it is critically important to have seamless processes  1. Data 2. Models 3. Output. Unfortunately, most of ML services are provided as an independent one from other services, therefore users should gather data and inform results of data analysis of stakeholders and management, one by one, independently, outside ML services. However, when we see the portal of MS Azure, Machine Learning is built on as one of the functions in MS Azure.   So we can operate this ML as one of the processes in MS Azure. It is completely different from other ML services.  Then let me go to  MS Azure ML studio and look at major functions in details.

After creating ML working space, we can go to ML studio where experiments can be done by using Graphical User Interface.

1.  Data

More than 30 data sets,  for example census income data,  are set up in advance.  So beginners can start data analysis immediately for training.  It is good because they can concentrate on data analysis in MS Azure ML.  Data, which are analyzed, should be just dragged and dropped into experiment area.  So data can be handled with their intuition.  No need to read manuals in advance.

 

2. Predictive models

In ML studio, there are more than 10 predictive models for classification.  Logistic regression, neural network and SVM., etc. are available here.  Models for regression and clustering are also available. According to the documents,  more than 300 R packages, which are open source in R language, are  also available.  It is amazing that these models can be used by drug and drop in ML studio without writing code. So beginners can analyze data without coding the models.

 

3. Output

Once data analysis is completed and predictive models are developed,  it is easy to release it as web application services by clicking the buttons to deploy it in the web. It is usually difficult to explain how predictive models work  just by theory.  Web applications must be powerful tools to explain how the models work to stakeholders, managements and customers because web applications can show us the results based on inputs from users.

 

As I said before, it is critically important to have seamless process 1. Data  2.  Model  3. Output.  Microsoft Azure ML realizes this as a cloud service. I would like to develop interesting web services based on Machine Learning in the future. The current version of  MS Azure ML is a preview,  so functionalities might be changed or removed, added going forward. If you need more information about MS Azure ML, please refer to this web. Let us enjoy machine learning !

Advertisements

How can beginners learn computer programming easily?

binary-65473_1280

 

Machine learning as a service is started by Microsoft Azure.  It seems that we can analyze data without writing code.  Does this mean we do not need to learn the computer programming anymore ?  I do not think so.  Because computer programming is needed to add new functions and provide fine tune to existing statistical models.  This is like driving cars.  Although there is no need to do “gear changes” manually,  we can run faster with gear changed manually when drivers have skills.  In addition to that, learning programming is learning how the calculation is done on computers.  The more you learn programming, the more you can understand methods of calculations in computers.  So do not be away from computer programming.

For beginners, however, it is not easy to start computer programming. There are a lot of books and manuals about computer programming.    Unfortunately, most of them are written for engineers and programmers, rather than beginners. So I would like to consider how beginners learn computer programing with ease.

 

1.  Do not hesitate to re-learn high school math

First, we should consider how problems can be solved and results can be obtained.  It is called “algorithm”.  Algorithms usually are expressed by mathematical formulas.  Whether therefore you may like math or not,  knowledge about elementary math, especially vectors and matrix is needed in computer programing.  Data is stored by using “Vectors and Matrix”, so it is very important to be familiar with Vectors and Matrix in advance.  The calculus is also important and provided strong recommendations to learn.  Most of the knowledge needed can be obtained through high school math textbooks. No need to solve complex problems,  just read textbooks and understand how they work. With math, we can prove “algorithm” is right.  It is a shortcut to learn computer programming.

 

2.  Let us be familiar to manipulate “Vectors and Matrix”

Finally algorithm should be implemented on computers. This is the last step you which is overcome. In computer programming,  “X” does not mean one number such as “1”, “64” but group of  numbers “2”,”56″,”789″.  It makes presentations of data simple and easy if you can understand how it works.  So be familiar to present matrix using “X”, “Y” even though each of them looks one number. This is the key concept for beginners to understand how data analysis is processed on computers.

 

3. Exercise computer programming without a computer

You do not need to sit down in front of computers in learning computer programing.  Just a piece of paper and a pen are needed.  Let us write down the algorithm you learned before. If you understand it with 120%, you can write code from the beginning to the end correctly.  Anytime and anywhere we can learn computer programming using  a piece of paper and a pen.

 

Anyway,  do not be shy to computer programing and just start it step by step.  You might be an expert of this field in the future. Good Luck!!

Is this a real game changer in data analysis, isn’t it ?

PC 0193

Machine learning (ML) is considered to be difficult to be implemented in practice. However, it has not been difficult anymore because new service is available from Microsoft.  This service is called “Microsoft Azure ML“. It can change the way to use Machine Learning in business.  It can be a game changer in data analysis, too.

 

1.  ML is integrated in the seamless processes.

ML can not be exist independently. Data should be gathered and cleaned up, models should be generated and validated correctly and results of data analysis should be shared among corporate managements and used for making better business decisions. Microsoft Azure ML realizes this ideal environment. So it is a user-friendly tool and can be fit for beginners of data analysis. It can be a strategic management tool, too.

2.  We can use R.

R is a popular language in data analysis. Microsoft Azure supports the R language. Even though there is no need to write code  by ourselves to start analyzing data,  the code for the models written in R can be seen on the screen.  It is very good because you can do fine tune to the models if you have expertise on computer languages.  In addition to that, more than 350 packages written in R are available, too.  Therefore, we do not need to write the code of  models by ourselves.  We just use these packages, which are open source in order to analyze data.  If you try to develop your own models by using R, of course  it is possible.  So Microsoft Azure ML is also good for experts of R language.

3.  The cost is dramatically decreased to start data analysis by ML.

10 years ago,  proprietary software of data mining was too expensive for individuals and start-up.  But Microsoft Azure ML can be used with “Pay as you go”. So initial cost can be decreased dramatically.  It means that  application services can be developed even by individuals with less cost.  It must be wonderful for entrepreneurs who try to make the world better by data analysis.  I want to do that.

 

This service has just started as preview and only available in the US south central so far.  Whenever it is available in Asia, I would like to try this service.  When I write this blog,  I heard the news, which says Apple and IBM team up to expand cloud services in corporate business.  Therefore, this kind of services about data analysis based on the cloud may be released from competitors of Microsoft in the future.  It must be good for us as we can choose the ML services based on our needs.  In the near future,  I am sure ML dominates the digital services for customers to choose good products and services.  Let us start now!

My memory of bubble economy in Japan

DataHero Nikkei 225 (Max in the year)

Stock markets in the US are very active recently.  The Dow industrial average went beyond 17,000 and hit the record high last week.  Stock market in Kuala Lumpur is also active, FTSE Bursa Malaysia KLCI Index was about to reach the highest last week. It reminds me the bubble economy of Japan in the end of 1980s. It was definitely enthusiasm at that time. Stocks would be going up tomorrow because it went up today. This is a story in Japan,1989.

 

1. Situations at the end of 1989 in Japan.

Nikkei 225 hit a historical all-time high of 38,957.44 in December 29, 1989.  When I saw the screen board,  I wondered why it happened because I did not find the reason for Nikkei 225 to reach the highest.  In 1989, the stock market went up from approximately 30,000  to nearly 40,000. But I did not see any big news to justify that in 1989.   Without a economic mechanism changed,  I thought that just the only price of stocks had been going up.  It was strange, definitely.  At the beginning of year,  a lot of economists and think tanks predicted movement of the stock markets in the year,  I remembered one of economic research centers said “Nikkei 225 will reach 100,000 in the near future.” at the beginning of 1990.  However, current Nikkei 225 is around 15,000, as you know. What’s the difference it is!

 

2.  How it happened.

In 1989, I worked at one of the four biggest Japanese investment banks as a stockbroker.  My main clients were retail investors, such as owners of small companies,  doctors,  professionals and housewives!  There were a lot of economic research to analyze Japanese economy and  justify this crazy appreciation of the stock market in Japan. But the reason why it appreciated was very simple.   The stock market went up because everyone bought it.  I talked a lot of retail investors everyday.  Beginners are usually quite conservative at first,  however once they realized every stock was going up and most people around them bought stocks this year, beginners also start investing aggressively although they had little experience.  It is a little difficult to know how it works because it is not a normal situation.  Can you imagine every housewife starts investing stocks today?   In 1989, they did.   I heard this phrase many times in 1989,  ” He does not invest stocks as he is foolish”.

 

3.  What happened after bubble burst?

It was miserable after the bubble burst as you know.  Most investors lost their money and only debts remained.  A lot of companies, including listed companies went out of the markets.  The most important thing,  I think, is that Japanese people lost confidence to go forward with taking risks.  Emotionally, this impact was strong enough to discourage the economic growth.  Everyone in Japan felt to get richer and richer until the bubble burst, but suddenly realized it was a bubble and their wealth disappeared from sight.  This is the beginning of a lost decade in Japan.

 

It is said that human’s memory is short-lived, so bubble will appear again every 20-30 years.  I would like to monitor when it happens again.

Credit Risk Management and Machine Learning in future

macbook-336704_640

Credit risk is always s hot topic in economic news.   Are US Treasury bonds safe? Does the Chinese banking system have problems?   Is the ratio of household debt to GDP is high in Malaysia?  etc. When we think about credit risk, it is critically important to know how we measure credit risk. So I would like to reconsider how we can apply Machine learning to credit risk management in financial institutions.  When I used to be a credit risk manager in a Japanese consumer finance company 10 years ago,  there were no cloud service and no social media. However, now that it is the time of big data,  it is a good timing to reconsider how we can manage credit risk by using the latest technologies such as cloud services and Machine learning. Let me take three points as follows.

 

1. Statistical models

One of the key metrics in credit risk management is the probability of  defaults (PD).  It is usually calculated from statistical models such as regression analysis.  Machine learning has algorithms about regression analysis, therefore it must  be easy to implement Machine learning to credit risk management systems in financial intuitions. Once PD is calculated from Machine learning,  this figure can be used the same way as current practices in financial institutions.  Statistical models usually have versions since it is developed but there is no need to worry about version control because it is easy on the cloud system.

 

2. Data

Data is more different from one which used to be 10 years ago.   I used private and closed data only in order to calculate the risk for each borrower.  Now that social media data and open public data are available in risk management,  we need to consider how to use these kind of data in practice. For this purpose,  cloud services are good because they are scalable and easy to expand the capacity to store data whenever it is needed.

 

3.  Product development

Recently product development is fast and active to keep the competitive edge as customer can choose channels, to contact with financial institutions and are getting more demanding.  That is why risk management in financial institutions should be flexible to update product portfolio and adjust methods in the light of  characteristics of new products.  Combination of cloud service and machine learning enable us to develop the risk models, enough to quick to keep up with new products and change of business environments.

 

Unlike retail industries,  financial industries are heavily regulated and required to audit risk management systems periodically. Therefore, audit trails are also important when Machine learning is applied to credit risk management.  I think combination cloud services and machine learning is good to enhance credit risk management with cost-effectiveness  in the long run. I would like to try this combination in credit risk management  if I can do it.