https://twitter
THREAD RE DOMINION SERBIA
"I was researching companies to see how many IT people they employ in US vs overseas, and I found a feature which lets me search LinkedIn by region, so I thought I would apply the same to DVS and see if we can contact any developers in Serbia to get them to flip on DVS".
"I found about 10 people working in Serbia for DVS, and I started to look into the developers, I started to notice that the developers were linked to the company, but more than a few did NOT have the DVS experience listed on their profile"
"How does that happen? It happens when you used to have the experience linked but it's since been removed. LinkedIn Caches those employee employer relationships so you stay listed under their company even though you've removed the experience."
"That got me wondering WHY would someone want to remove their relationship with the company… only answer is to cover something up. So I started to research each one.
Here is the screenshot of the DVS Employees on LinkedIn:"
"You can see on this screen shot that I am hovering over Aleksandar Lazarevic and that the link to his profile is showing in the bottom left of the screen. The url being: ( https://linkedin
"Note the handle ending in 777, that will come into play later.
In researching this developer I discover this person wrote the book on how to do algorithmic boosting of datasets over time. The Official name for it is SMOTE (Synthetic Minority Over-sampling Technique),"
"I've attached the paper to this email and it's still currently available online here ( https://www3
"I've attached the paper to this email and it's still currently available online here ( https://www3
"Unlike standard boosting where all misclassified examples are given equal weights, SMOTEBoost creates synthetic examples from the rare or minority class, thus indirectly changing the updating weights and compensating for skewed distributions"
"SMOTEBoost applied to several highly and moderately imbalanced data sets shows improvement in prediction performance on the minority class and overall improved F-values."
"Layman Description: An adjustment will be made based on the top data point to keep the bottom data point on track to balance, or Tie. Now apply that idea to tabulating votes."