### 08-13, 11:15–11:55 (Asia/Yerevan), 114W PAB

"Prediction is very difficult, especially if it’s about the future!" This phrase is attributed to Niels Bohr, the Nobel laureate in Physics and father of the atomic model. This quote warns about the unreliability of forecasts without proper testing and about constant changes in the initial assumed conditions.

With modern programming languages and convenient packages that provide ready-made modeling solutions, it is often easy to find a model that fits the past data well; perhaps too well! But does the maximization of metrics justify the means? Should the complex structures of predictions be built on the quicksand of noisy data?

This talk is a laid-back discussion that will be useful for the audience from any background, from beginner to advanced. Aghasi Tavadyan is the founder of Tvyal.com, which translates to "data" from Armenian. You can find more info about him following these websites: tavadyan.com, tvyal.com.

It is relatively easy to write a few lines of code and declare a black-box complex model that maximizes the metrics. This is often done without understanding the underlying assumptions. But do those complex models justify the means? The models with the highest metrics may be good for competitions, but everything has a tradeoff. Complex models with maximized metrics may have unwanted edge cases, be hard to explain, and hard to modify.

In the real world, it is another matter to find a model that correctly identifies those features of the past data which will be replicated in the future. The model that fits well, will not predict well. One should always remember that models are built on assumptions and noisy samples and the underling game-rules of the phenomenon we want to predict may constantly change.

Modeling and forecasting should not be the mindless rituals of hypothesis testing or applying the same tools to the different data sets. Each data set is unique has its own assumptions and the level of noise. Statistics and machine learning provide a huge toolbox. We should not blindly use only a few tools for every situation. We should always be inquisitive about the given data.

**Prior Knowledge Expected**–

No previous knowledge expected

Aghasi Tavadyan is the founder of the data science organization tvyal.com, which translates to "data" from Armenian. Tvyal is data analysis, modeling, and visualizations. It simplifies the complexity of any data-blobs to digestible pieces and forms the T-wings that can help your business fly more effectively. He is the head of "Economic Uncertainty Modeling" laboratory, ASUE, and the executive secretary of "Armenian Economic Journal," which is published by the National Academy of Sciences, Armenia.

Aghasi Tavadyan teaches Probability Theory, Mathematical Statistics, Econometrics, Mathematical Models in Financial Markets, Bayesian Statistics at the Armenian State University of Economics and Armenian-Russian Universities.

Please visit his websites for more info: tavadyan.com, tvyal.com