Year of Graduation
Risk modeling of innovative project using neural technology
Faculty of Economics
One of the main problems of the Russian economy is the slowdown in innovation. Particularly low levels of innovation activity is the construction industry. One of the main reasons for this are the high risks that accompany any innovation activities (including construction). In this regard, the important question is risk modeling innovative projects. One of the most common methods of modeling of economic risks of any projects are neural (neural network) technology. The purpose of this master dissertation is the modeling of the risks of innovation project "PROMTEKHSTROY" using neural technologies. Achievimg this goal we need to solve the following objectives:- Define the essence of the innovation project,- Explore the concept and the types of risks in the implementation of innovative projects- Explore how we can use technology to simulate neural risk innovative projects-To apply this method for risk assessment of the innovative project "PROMTEKHSTROY"- Assess the risks of innovation project alternative method (Monte Carlo) and compare the results. This master dissertation includes technical and economic characteristics of the innovative project "PROMTEKHSTROY", sensitivity analysis, scenario analysis and simulation Monte Carlo. In this paper, there was built the optimal neural network’s model, which architecture has [7-2-1] and 19 synaptic weights. The result of the sensitivity analysis is the following: to innovation project "PROMTEKHSTROY" was successful. Further analysis was carried out scenarios in which received the value NPV and "optimistic" scenario (the best) and "pessimistic" (worst case) is positive. Therefore, the study should take an innovative project to implement, it will be profitable. Correctness of the previous output and simulation confirmed by the Monte-Carlo scatter predictive values NPV is in the range [48459.528; 40508.092]. Standard deviation = 877.09. Most probable value of NPV equal received 40,483.57 thousand rubles. Risk modeling of innovative project using neural technology got the next result: predictive value NPV equals 40,927 th.rub, Calculated that more (38840 th.rub.) and obtained by the Monte Carlo method (40836,57 th.rub.). It should be noted that the standard deviation of the predicted value NPV, obtained using neural technology equals 121.8, that in 7.2 times less than that obtained using the Monte-Carlo method. Error training and testing the neural network is negligible, indicating that the high quality of the constructed model and the possibility of further applications for similar purposes. The result of this master thesis is the conclusion that the unit of neural networks can be successfully applied for innovative projects subject to the availability of training sample. Thus, companies with data on the implementation of innovative projects past, can safely use neural technology to reduce their production risks. However, despite the high accuracy of the results of neural network modeling, we should not forget that for the adoption of adequate solutions for an innovative project, it is advisable to use both methods.