Can Machine Learning solve today’s problem? Maybe yes, because…



Machine Learning is a hot topic this year.  Machine learning is defined as follows,  Field of study that gives computers the ability to learn without being explicitly programmed, Arthur Samuel (1959).  As a lot of data and computer resources are available with less costs recently, Machine learning is getting popular in the field of data analysis.

In academics, there is no doubt that Machine learning has a good performance in statistical computing.  Then how about the real world?  When we try to apply Machine learning to data analysis on a daily basis,  there are two difficulties to cope with. One is making data sets for training and the other is implementations of the models on computers in order to obtain the results from the models. As the name of  ‘learning’ suggests, the training data set is required so that computers can learn the data before models generate the results from observed data. In order to implement these processes,  knowledge and expertise about data analytics are required to complete the tasks. It must take one week or one month depend on the availability of resources of data scientists.

So I thought it might be difficult to solve today’s problem within today before I have heard the announcement from Microsoft on 16 June 2014.  This is about Microsoft Azure ML, which is Machine learning statistical tool operated on its platform “Azure“.  Azure is one of the platforms on the cloud.  So it competes with Google apps for business and AWS.  Although the details is not disclosed yet,  it looks like that Azure ML is better than other analytics tools in order to establish seamless processes from preparing the data set to model implementation on computers,  because Azure itself is a seamless process platform and Azure ML is a part of them, not exists independently. So users do not need to pay attentions to the relationship among each independent component in the platform.  All they have to do is just to consume data and obtain the results.  It means that we can go short cut in analyzing data on a daily basis.  It is critically important because the quick response against change of business environment is required in data driven management. Microsoft says that the preview of  Azure ML will be started in July.  It must be exciting and must enhance the data driven management.

Once Machine learning is getting a user-friendly tool,  what should business managers do?  I think  it is very important to realize what data around us is available and will be in future.  Data is a starting point for data analysis and data which is available has increased exponentially.  ‘The data we create and copy annually’  is doubling in size every two years  from 2013 to 2020 according to the research conducted by IDC.   Yes, we should be data savvy managers as Machine learning stands by us! Continue reading

Is it a new star in MOOCs ?


Followed by last week’s blog,  I would like to write MOOCs again because I found the potential new star in MOOCs yesterday.   This is “Nanodegree” provided by Udacity, one of the big names in MOOCs.   Unlike Edex and Coursera,  two big names of MOOCs,  Udacity have a close relationship with industries, such as AT&T,  Facebook and Google.  Therefore “Nanodegree” courses will provide project based programs and make participants be ready to be hired.  According to the blog of Udacity,  AT&T is offering 100 paid internships to top graduates of the nanodegree program and will consider students with nanodegrees when there is a potential job match. Nanodegree is a trademark of Udacity.


Comparing other courses that I took before in MOOCs, nanodegree might have advantages to them.

1.  It is focused on “being ready to be hired”

It is always said that current higher education cannot meet the demand from industries so there is a gap between knowledge and skills of college graduates and ones required by employers.  Although the details of the courses are not disclosed, it is expected to be based on activities and operations in industries, rather than academics.  From the standpoint of employers,  it is good that Nanodegree is created for the needs of industries, so that It has far less mismatch to the needs of the industries.

2.  It is not free, however, still far cheaper than on-campus courses.

It will cost $200 USD per month and it takes from 6 to 12 months to complete the courses.  Compared with the on-campus courses, it is much cheaper.  I think one of the reasons not to be free is that coaching is available to participants during the courses.  It is considered to be necessary for participants to learn the courses effectively and be motivated through the courses.

3. It takes less than one year to finish the courses and obtain the certificates

According to Udacity,  it takes from 6 to 12 months to complete the courses when participants use 10 hours per week for the course.  So it may be completed less than 6 months when participants use more than 10 hours per week to learn the course.  It is very good, especially for participants who are already working.  Although course sequences are also available In Edex and Coursera (X-series in Edex, Specialization in Coursera),  some of them can not be completed within 6 months. Business environment changes so quickly so the shorter the course is,  the better it is for participants.


According to Udacity blog, nanodegree is expected to start this fall in 2014.  There will be five courses, front-end web developer, back-end web developer, iOS mobile developer, Android mobile developer and data analyst. I would like to take one of them and obtain the expertise of it. It must be exciting!


Ultimate knowledge source “MOOCs”



Recently I took several courses in MOOCs and enjoyed them.  Do you know “MOOCs”?  MOOCs is massive open online courses.   A lot of famous universities  provide us their courses from computer science to Art & Literature.  Edex, Coursera and Udacity are famous brands in MOOCs.   They started providing courses in 2012 and now millions people take courses through their platforms.  It consists of videos,  documents,  questions and mid/final exam. When you obtain the score over the criteria which depends on courses, you can get the certificate.  I was very surprised when I knew most of the courses are provided as free.  When you want the certification with your identification, it takes reasonable costs.  But courses itself are basically provided as free.  So MOOCs is expected to lower the barrier to enter higher educations in developing countries.

Currently I am taking two courses about credit risk management by TU delft and financial programing and policies by IMF in Edex.  They are very good to brush up my knowledge. Each author of both courses is a professor in one of top universities or international institutions,  contents are updated with the current situation.  Both of them are strongly recommended.


Since last year, I have taken more than 10 courses in MOOCs.  Based on this experience, I can explain what the pros and cons of MOOCs are.


1. Most courses are new and updated periodically.

This is the advantage of digital contents and the internet.  Most courses are updated to react the performance of previous versions of courses. More than 1000 or 10000 people participate and provide their opinions to the courses so it is always improving steadily.

2. We can take courses anytime and anywhere

This is also the advantage of digital contents and the internet.  As long as you have an access to the internet, you can enjoy MOOCs anytime and anywhere  all over the world.  I moved from Tokyo to Kuala Lumpur in the middle of courses this year, but there is no problem to continue it.  This is very good when jobs are becoming mobile and cross border.



1. It is very difficult to keep your motivation to finish the courses.

There is no teacher in the classroom physically when you take MOOCs.  There is no warning for you to skip one of learning sequences or miss the exams. Although there are  a lot of participants, only less than 10 percent of them finish the courses in MOOCs.  It is a problem that we should solve going forward.

2. There are a lot of courses so it may be difficult to choose the best one for each participant.

As a lot of universities have participated in MOOCs recently.  You can see a lot of courses in the certain topic, so it is getting difficult to find courses which fit your learning needs in MOOCs. In nature, we cannot see the whole story of the courses in advance (we cannot check each video in advance), although the course summary is available.  I sometimes refer participants reviews and opinions to the course before I take it.


Anyway, I love MOOCs and continue to learn their courses.   Next week I will start the course about machine learning by the famous professor of Stanford University.  I hope I will enjoy it and learn machine learning a lot!  I would like to report how this machine learning course is going.

Thailand from the past to the present


Investors and international business people worry about what will happen in Thailand. After the coup happened,  it seems that no solutions of this turmoil can be seen in the near future.  Here I would like to look at some data of Thailand and help us to understand what Thailand is.


DataHero Urban Population By Country

This chart shows how many percent of the total population lives in urban areas by county.  In Thailand, around 35 percent of people live in an urban area in 2012.  It means that nearly 65% people live in a rural area in Thailand. This rate has increased only gradually since 2005.  Compare with other Asian countries, the concentration of population is moderate in Thailand.  Most reports of the mass media are broadcasted from Bangkok. However, 65% of people live in a rural area. So I would like to see what people in rural area think of the coup.


DataHero Internet Users By Country-2

This chart shows how many people use internet out of 100.  In Thailand the ratio is around 26.5 in 2012.  It seems that internet users are not increasing as Singapore and Malaysia in terms of ratio against the total population.  There is another statistic. According to data from Social Inc., as of last month, 28 million Thailand users are on Facebook, 4.5 million have joined Twitter and 1.7 million has Instagram accounts, it means more than 40 percent of the total population is on Facebook. I think a lot of user start using the internet in Thailand recently.

Not only in Thailand but also other countries which limit democracy,  the first step by the governmental body is to control the mass media and the internet in the country. Although the penetration rate is not so high in Thailand compared with other countries, Military of Thailand may control internet traffic in the country.


DataHero International Tourism Arrivals By Country-2

I am wondering how the tourism industry in Thailand is suffering from this turmoil.  Since 2009,  Tourists from abroad has been increasing steadily in Thailand.  So if  this turmoil and instability continue, Tourists from abroad must be decreasing.  It is very bad for the tourism in Thailand.


I presented only three data about Thailand and other countries here.  The more data are presented,  the better we can understand what the country is.   Those data are obtained by Quandl, the data provide service. If you are interested in data about countries, please go to the company’s web site.  You can find a lot of information in it. Once you obtain data that you are interested in, you may like to visualize them. I recommend DataHero to do it as it is very easy and efficient to do so. The charts above are created by DataHero.  Is it cool, isn’t it?