The Moscow State University supercomputer, Lomonosov, predicted the social and economic development of Russia for the next 50 years. The project goal was to acquire the experience of building such social and economic simulation models in Russia. The project team was comprised of two specialists from the Central Economics and Mathematics Institute (CEMI) and three specialists from Moscow State University (MSU).
The model included such factors as population number change in particular regions and in the whole country, GDP dynamics, amount of investments in industry, and country and innovation industry EVA change.
200 supercomputer processors were used in the evaluation. The whole process took only 1 1/2 minutes. The model was also tried on 1000 processors, and then it took only 16 seconds. In addition to Lomonosov, the evaluation was also carried out on supercomputers MVS-100K (Joint Supercomputer Center of Russian Academy of Sciences) and Chebishev (MSU). Regular computers are unable to calculate such models (top productivity of Lomonosov after its modernization is 510 TFLOPS).
Modeling showed that in 50 years, the north territories’ population will be almost completely disbanded and the Siberian and Far East populations will significantly decrease. South regions are expected to meet population growth. According to the model, GDP on the whole is expected to grow. Yet, the specialists who took part in the project warn not to take the results as an inevitable forecast, as the model did not take into account such possible influential events as war or a huge epidemic.
The Russian social and economic development model was developed with use of simulation software AnyLogic using Agent Based modeling technology. To create this model, CEMI specialists used about 100 million agents. The data used in the simulation was taken from the Federal Agency of State Statistics and from Russian monitoring of the economic situation and the health of the population.
Getting the final results, the scientists say, wasn’t the only objective. They also wanted to get experience using model conversion and transferring it from a regular PC to a supercomputer because there is a lack of such experience in Russia. One difficulty was to parallelize Agent Based model parts to different computational nodes of the system. Another complication was the technical problems involving model transfer to supercomputer program code. These were the tasks of the MSU specialists.
Large scale social and economic models were also created by AnyLogic users in the USA, Germany, and Sweden. For example, RTI International, a research institute in the USA, used this software to simulate HIV/AIDS proliferation via drug addicts. Also, the United States Census Bureau worked with AnyLogic using System Dynamics modeling for the Hispanic population in California.