Teaching at AIMS Data Science Workshop

I had the privilege to be able to present a number of topics at the African Institute for Mathematical Science's Data Science workshop. I really enjoyed my time teaching, learning from the other speakers and meeting very talented students from around the country.

I presented 3 talks covering

  • Applications of Data Science to Social Good
  • Unsupervised Learning and Reinforcement Learning (Slides)
  • Visualisation (Slides)

Posted in CSIR, Data Science, Work Tagged with: , , , ,

Data Science #1: Science Stars Article

First article in the DST Magazine ScienceStars (Aimed at informing youth on science matters)

You can read it here.

Posted in CSIR, Data Science, Work Tagged with: , ,

A dive into Word2Vec + TSNE for visualisation.

This is a quick post. Spent some of Human Rights day playing with Word2Vec on a dataset we have of Tweets collected from Gauteng for a 2 week period in February 2017. You can see the results of Gauteng Twitter Word2Vec visualised in 2D using tSNE below.

Gauteng Twitter Word2Vec visualised in 2D using tSNE

Gauteng Twitter Word2Vec visualised in 2D using tSNE

Nice thing about this is that you can look at clusters of words that relate to a concept. This cluster is of Football

Gauteng Twitter Word2Vec Cluster [Football]

Gauteng Twitter Word2Vec Cluster [Football]

Gauteng Twitter Word2Vec Cluster [People Description]

Gauteng Twitter Word2Vec Cluster [People Description]

Gauteng Twitter Word2Vec Cluster [Local Language Exclamations]

Gauteng Twitter Word2Vec Cluster [Local Language Exclamations]

Posted in Ramblings, Technical Hacks Tagged with: , , ,

Road Traffic Fatal Crashes

This a developing posts.

  • Version 0.5: Initial post.

My SO and I have been having a lot of conversations about Fatal Crashes in South Africa. It is a very popular issue to discuss in South Africa especially during the festive season. To start answering some of the questions about the impact of the government and law enforcement on reducing the crashes, we first have to gather the data and start looking at it. To that end I jumped into reports from RTMC. I have made available a simple dataset of Fatal Crashes per South African province available. Here is a plot with the top 7 provinces

Fatal Crashes (Top 7 South African Provinces)

Fatal Crashes (Top 7 South African Provinces)

Data

You can get the data from my Github. It's available here https://github.com/bionicv/datadriven/tree/master/fatal-crashes. You can load this file into GapMinder Offline.

GapMinder SA Fatal Crashes

GapMinder SA Fatal Crashes

Posted in Data Science, Technical Hacks, Travel, Work Tagged with: , ,

So you want to be a MMM superstar?

Original edited post appeared on iAfrikan [Evernote Mirror]

What if I offered you a great opportunity to get over 30% return on investment, every month. Yes, you read that correctly 30% every month. That is about 2,240% interest per year. So if you gave me R1000 at the start of the year, I should give you back R23,400 at the end of the year. This sounds too good to be true right. We’ll it is. It should get to you to start questioning, what actually am I doing with the money so that I can give you so much interest. Have I started some company that is experiencing guaranteed hyper-growth for all this time. Did I discover oil in my backyard, or maybe sit on land that has a large gold deposit in my community.

A lot has been said and written about MMM (or KIPI or insert your newest scheme here) all over the world. In South Africa, it quickly gained fame and infamy over the last few years. It’s questionable legal status has also allowed to take advantage of grey areas to be able to continue to advertise itself while regulators try to figure out how to classify such schemes. There are obvious red flags. Promising large monthly returns is the biggest. I does not matter how the scheme tries to dress it up using terms like “donating”, “giving help” etc. All of these tend to distract from the core matter. Secondly, using the argument that it is the same as a group savings scheme (that is a Stokvel) is also dishonest.

If you are in a stokvel, let’s say with 11 other people, contributing R1000 a month, then the maximum possible monthly circulation of money is R11000 if the scheme uses a monthly payout to a single member of the stokvel. Every member’s payout is R11000 (R12000 - R1000 from the members own potential contribution). The members cannot get more money, than as stated, from the scheme as it would require someone to give more than their R1000 monthly contribution. Even if we were to change the scheme to now only pay out members every December (#KeDezembaBoss) the same challenge arises. The Stokvel could be more forward thinking and have a medium growth investment account where they deposit their money's monthly. Even with favourable conditions, let’s say they get 6% growth on the account (reference) they would have R148,026.75 from total deposits of R144,000.00 (12 members * 12 monthly * R1000 monthly contribution). As such they would have an extra R4,026.75 just from interest from the whole first year. So long story short, a 35% monthly return using an investment account is also out of question.

So how doe MMM/KIPI etc. do it? It’s simple, it’s a Ponzi scheme. In a Ponzi scheme, high returns are paid to older “investors” using contributions of new “investors”. As there is no investment vehicle or real business that is taking place to actually get revenue from the deposits, money just circulates as it is being depleted (I referred to how you cannot really get more than you put in, if so, someone has to be contributing more or losing money). What makes your MMM/KIPIs aggravating, is that their scheme is quite clear that this is what they are doing. It is simply not sustainable. The only way to keep payouts going is to aggressively be signing up more people. The earlier “investors” get some payouts and are heavily encouraged to give testimonials, while other payout schemes are created to give bonuses to those who get others to signup.

The more signups you have the more you need to keep payouts going, very soon you run out of people willing to play this game and the schemes start going through “freezes”. Lots of jargon will be written about the reasons of these freezes and how there might be third forces causing the freeze (Note: I already expect loyalists to the schemes to send lots of nice comments our way so I will spell things out clearly). All those jargon rich explanations are just a smoke screen, the scheme has started collapsing, if it has not already collapsed.

Now you might ask if money is circulating between members, how might the schemes owners make money? There are many possible ways. KIPI for examples asks for a subscription (about a R1000 a year). Let's say they had 500,000 active members. That is R500 million rand yearly just for members to play the game. They are likely not using any of this money to give out that 35% growth they promise. Just money in pocket. For MMM, it might look like it’s transparent but there is no guarantee that the accounts that other members deposit into (termed Giving Help) are not the scheme owners themselves or conduits to them. They could simply pocket 5% of all deposits and still have a comfortable cushion. It would make their scheme even more unstable, but you just need bursts now and then and you are making lots of money.

Earlier in the year, after learning of these schemes, I offered publicly to personally help people who are interested in this scheme by offering them a guarantee of whatever money they would invest, if they invested in me, I would give the 50% back. So if you gave me R100, I would guarantee you will get back R50. As such a loss of 50%. The reason being that I thought a loss of 50% is better than a loss of 100%. It was not me creating a bad omen on the schemes, it is already written on the wall at the start of the scheme. People will lose a lot of money. What goes unsaid is how bad the losses are and the personal losses.

I took some time to look at the scheme’s Facebook groups (some closed/private) to understand what was going on and was going unsaid. A lot of  people venting frustrations on their money being “frozen” in these systems, aggressive leaders (those likely getting bonuses to recruit) encouraging members to keep throwing more money in on the promise that the schemes will “restart” and other members selling a 100 other similar schemes. Basically it’s just a mess. The groups that look very subdued likely had an administrator who removed all content that might look negative on MMM/KIPI. So everybody who is critical, is likely going to get banned and their posts removed. People have put in their life savings, have taken out loans and have gone without a lot on the promise of high returns. Yes these schemes are not new and more will pop up in future. And it is not about labelling people as being stupid to fall for them. We should have compassion while at the same time exposing the schemes themselves.

Being a researcher around the Information Security space, one of the biggest concerns I have is the amount of personal data these schemes have that is not under some regulatory oversight or protection guarantees. Are these schemes making further money by selling your data to unscrupulous 3rd parties? How are you going to be protected when their databases with your account numbers etc. get leaked? Who do you go to if you are further scammed by the 100s of other scammers using the system? These schemes could also be selling your bank details to organisations running those R99 debit order scams.

The basic truth here is that these schemes appeal to the greed in all of us. We all would like to get high returns for not much work so we can have a more comfortable life. Hiding the losses and lives broken becomes part of us saving face so as to not look like an embarrassment to others. But, in cases like this, it is better to speak out about the losses so as to warn others who want to get into these schemes. Your MMMs/KIPIs will likely not be forced to publish data on what the average member makes. like how Multilevel Marketing (MLM) companies like Herbalife/Amway are required to in some US states. For your information a majority (99.7%, so using majority generously here) of these MLM “distributors” make losses. Where we go from here, I do not know. I do hope with this article I add to the conversation.

Posted in Everthing Else, General, Ramblings Tagged with: , , , , , , ,
Tweeter button Facebook button Linkedin button