“RG” asked the headmaster, Ilya Mirin, how the data relate to the experts and that are still able to calculate the machine.
Ilya G. why the school decided to do with this topic?
Ilha Mirim: We participated in the competition, which was organized by the international community of Open Data Science, Artificial Intelligence Sberbank and Sberbank. The source data for programmers were confirmed cases of infection COVID-19 based on the research of Johns Hopkins University CSSE. Several hundred participants have built algorithms for the most accurate prediction for the next seven days, and at the end of each period the solutions were tested. We put several teams and eventually hit the top ten.
And then we began to wonder what forecasts of such algorithms can be used yet, because the question is very topical – as we will live further.
how?
Ilha Mirim: To start a small educational program. Projected number of cases? We take data on the number of cases COVID-19 countries and dates for the previous few days and offer the machine to process them. This is not an algorithm, the principle of which we understand, namely training. There is data on the input and output, and, many times repeating a simple operation of selection, in the end we get the result. The more data and more precise than they are, so it is closer to reality.
But it is important to consider a few things: such predictions work well in specific cases. In addition, they are based on what has already taken place, but if something unexpected occurs – war, strong-willed decisions of the government, the machine is unable to predict the development of the situation. Accordingly, the future of the economy it is already possible to predict because the recovery processes in different countries started and has data to use.
We decided not to take global economic indicators – GDP, GRP, and the like, because the lives of ordinary citizens they are not particularly influenced more interesting to economists than ordinary people.
what are ordinary Russians?
Ilha Mirim: There are basic things that most people associate their well-being. For example, the cost of gasoline – the first indicator. We took the 95th, because it has the engines of Lada cars and Renault, which go for many. Yes, in the far East the situation is different, but according to statistics, most Russian regions the way is still AI-95.
the Second indicator is the purchase of new vehicles. It is reasonable to assume that this is a purchase that a person makes with the stock of money. The market is quite large – in the peak years in Russia, sold to two million new vehicles in crisis level falls, so that it can be clearly seen the economic situation.
Third – the purchase of smartphones, one of long term use goods, which has become a necessity. Studies show that often a person first buys a new gadget, new clothes, what’s more, 25 – 30 percent smartphones sold on credit.
In the initial phase, we took several sources and verified them with the official statistics, see how they are joined together and which are of the dispersion, to give the machine more close to reality figures.
What is the forecast issued a neural network?
Ilha Mirim: We are seeing a decline in new car sales – 300 – 400 thousand a year in Russia. It was correlated with the forecasts of economic development. According to the Ministry, in January, it sold 105 thousand cars from showrooms in February – 126, March – 176, and in April, when there was a strong panic, people sat in isolation, 38. It should be noted that our estimates practically coincided with the data. The economic development Ministry predicts a significant decline in sales in the second quarter compared to the same period last year, which, in principle, expected. The June figures will appear in July, and we will be able to compare them with our predictions and also to assess whether the right methodology was chosen for forecasting.
According to our undergraduates in the coming months will grow by one to two percent of the price of gasoline, they can reach 47 rubles per liter of AI-95. The dollar is projected at up to 75 rubles. Smartphone sales will increase by about two million units. The turnover of the online retail will increase by 8 to 12 percent.
we Have the forecast until around the end of the year, but it will be adjusted as the countries of restrictions.
what else can your electronic expert?
Ilha Mirim: as for predictions, you can remember the program providing housing for orphans, which in the previous five years, Primorye has received little attention. The omission led to a situation where massive financial injections, begun now, will not help to solve the problem quickly. In the beginning of the year in Primorsky region was 7.8 thousand persons in this category a year earlier and 400 less. About 17 percent of the waiting list are concentrated in Vladivostok. At the same time in mid-February in the regional center, there were only 611 of apartments that fit the parameters of the program – a certain area, year of construction, with a specific value “square”.
Given the growing number of orphaned children, the rate of appearance on the market of suitable housing and other parameters, the neural network concluded that if the conditions remain the same, this “housing problem” will be solved by 2045. Just add money is not an option, because in our case, this can lead to ledthe increase in the cost of housing, and we get another problem – the dissatisfaction of the citizens. A number of management decisions. By the way, is the far Eastern regions, for example, Sakhalin oblast, where he successfully cope with this task.
Actually, we did a lot of state and municipal projects, it looks like we have a new direction in the analysis of state and municipal data. I think it will develop. This is facilitated by including in the current FEFU educational program Chief Data Officer, the basic idea of which is that every official is now primarily an engineer working with data, and its purpose is to calculate the problem and the cost of its solution.
in addition, we are keen to continue to engage projections tied to the level of living things, to make them as accurate as possible.
