تعیین قیمت بهینه با استفاده از الگوهای انحصار چندجانبه؛ رویکرد شبکه های عصبی و نظریه بازی ها

یکی از چالش‌هایی که صنعت فولاد با آن روبه‌رو می‌باشد، اداره و مدیریت زنجیره تامین می‌باشد. بر این اساس در تحقیق حاضر بر اساس سناریوهای 3 گانه عدم همکاری و حرکت هم‌زمان (کورنو)، عدم همکاری و حرکت ترتیبی (استکلبرگ) و رفتار همکاری (تبانی)، در زنجیره تامین فولاد پرداخته خواهد شد. روش تحقیق ازنظر هدف کاربردی می‌باشد. بازه زمانی تحقیق داده‌های فصلی 2011 تا 2020 و نرم‌افزار مورداستفاده نرم‌افزار متلب می‌باشد. در این مقاله یک مدل ترکیبی بر اساس شبکه‌های عصبی مصنوعی و تئوری بازی‌ها ارائه شده است تا بتواند در تعیین سطح قیمت و تولید بهینه به فعالان صنعت فولاد کمک کند. جهت پیش‌بینی قیمت فولاد از سه شبکه عصبی بیزین، بردارهای پشتیبان و پاد انتشارگراسبرگ بهره گرفته شد. نتایج بیانگر این واقعیت است که مدل پاد انتشار گراسبرگ دقت بالاتری در پیش‌بینی قیمت فولاد دارد. نتایج بیانگر این واقعیت است که با حرکت از سمت بازی کورنو به سمت بازی استکلبرگ و از بازی استکلبرگ به سمت بازی تبانی در زنجیره تامین موجب افزایش قیمت در صنعت فولاد به ازای هر تن 6 دلار و میزان عرضه محصول در دامنه 1500 تا 4000 تن خواهد بود، به عبارتی با افزایش سطح تبانی در بازار فولاد میزان محصول بیش‌تری در بازار عرضه‌شده و هم‌زمان سطح قیمت محصول نیز افزایش خواهد یافت که این امر موجب کاهش رفاه مصرف‌کننده فولاد در بازار خواهد شد.

determining the optimal price in the steel industry using multilateral monopoly patterns with the approach of neural networks and game theory

introductionthe field of supply chain management has focused on crucial topics such as coordination, cooperation, and competition among chain members. game theory has emerged as a valuable tool for examining supply chain management issues, as it analyzes various situations and their impact on supply chain performance (naimi sediq et al., 2013; shafi’i et al., 2018). while every action and performance within the supply chain influences the outcomes of the game, it does not solely determine them. the goal is to balance the parties involved in the supply chain and create satisfaction for the end customer (metinfer et al., 2018).although extensive research has been conducted in supply chain management within the steel industry, the impact of sanctions on nash equilibria and the application of three approaches (cournot, stackelberg, and collusion) to achieve game balance in different scenarios have not been thoroughly investigated. this research aims to fill this gap by addressing the mentioned research question. the current study focuses on determining the optimal price using an intelligent decision-making system based on game theory within the steel industry, considering the presence or absence of the sanctions variable.our country currently possesses several relative advantages in terms of steel production conditions, including abundant and affordable energy, iron ore and refractory raw materials, considerable steel production experience, and a skilled and cost-effective workforce. by acquiring new production technology, these advantages enable our country to play a competitive and influential role in the global steel market. however, the steel industry faces significant challenges such as price fluctuations, extreme price disparities across regions, and delayed delivery due to a lack of efficient supply chain management. therefore, the main research question aims to provide a model that incorporates key variables, such as the supply and demand of final and intermediate products in the steelmaking industry and the future trends in production and quantity changes.research methodthis article introduces a composite model that combines artificial neural networks and game theory to assist stakeholders in the steel industry in determining optimal production levels and price levels. to predict the price of steel, three techniques were employed: bayesian neural networks, support vectors, and grassberg anti-diffusion. additionally, to address the issue of binary identification in the neural network, three different network structures were introduced: feedforward network structure, competitive network structure, and backward associative memory network structure.research findingsthe first scenario is the non-cooperative game (cournot model scenario) where each participant aims to maximize their profit and would not deviate from their strategy as it would lead to a reduction in their profits. the second scenario is the sequential non-cooperative game (stackelberg model scenario), in which two chains engage in a confrontation of the stackelberg game type. the leader’s goal is to determine the best strategy while considering all rational strategies that follower players can employ to maximize their income. this scenario demonstrates that the rate of price and profit increase is lower compared to sequential and cooperative game modes. the third scenario is the cooperative game (collusion model scenario). in this scenario, the allocation of profits among the cooperating members is crucial to ensure the stability of their cooperation. the grassberg anti-diffusion method exhibits higher accuracy due to its higher true positive (tp) and true negative (tn) values compared to other algorithms. additionally, it has fewer false positives (fp) and false negatives (fn) because a higher tp and tn indicate more accurate predictions in the tested dataset, while fp and fn represent incorrect predictions.