Svydenko A. Modeling of pricing in the retail market of petroleum products in Ukraine

The dissertation is devoted to pricing issues in the retail market of petroleum products in Ukraine. The main feature of the pricing in this particular market is their lack of transparency and price asymmetry. Therefore, the relevant tasks are market monitoring and constructing forecasts of behavior of retail prices for petroleum products. The cointegration approach has shown its effectiveness in regard to oil prices, however, under conditions of significant exchange rate fluctuations, the influence of each price factor on retail prices varies over time. This forces us to use neural networks to predict the behavior of retail prices. As a result of the combination of several approaches for the formation of input data, the allocation of threats as input parameters, it was possible to build a system of short-term forecasting behavior of retail prices, which takes into account the dynamics of pricing of each market participant. In addition to price indicators, there is significant set of nondeterministic factors, that forms price destabilization threats to the market of petroleum products. In order to recognize beforehand and formulate a list of actual threats, a machine learning based system of recognition of threats which destabilize prices in the Ukrainian oil market was proposed. The peculiarity of the approach is that the list of threat groups can be expanded and refined depending on the tasks set. In the oligopolistic markets, which are close to the retail market of petroleum products in Ukraine, the decisions of each firm also affect the profits of competitors. Therefore, firms react to the actions of competitors and in each decision firms consider not only a direct effect on their income, but also the effects of the reactions of competitors. The suggested model of behavior of retail prices based on neural networks does not allow to investigate the interaction in the market. Using the agent approach, we can develop a model that is sufficiently simple to structure and adequately represent the nature of economic objects studied in this paper. The agent model allows to focus not on the model as a whole, but on its individual elements, which allows knowing the behavior of each individual "agent", to investigate their interaction. We developed 2 agent models of the oil product market: under a collusion and on a consumer search basis. In order to increase market efficiency, the state can use both administrative and market mechanisms for market's regulating and stimulating it to establish effective retail prices. Such measures may include the following: use of the stabilization reserve; informing consumers to reduce information asymmetry; use of price regulation mechanisms based on the price corridor. Among the main factors of such a strategy of market regulation can be identified asymmetry of information on the market and the lack of state regulation of operational information about the relevant threats to the market of petroleum products. Creation of the information system of the market of oil products of Ukraine will allow to promptly implement appropriate measures to ensure the effective functioning of the market.