This article is focused on the creation of a system for forecasting of future orders for a specific beverage cans manufacturer. The problem comes from the irregular ordering of cans from different customers; not only national companies, but also from companies abroad. This causes fluctuations in production and consequently an irregular transport regime. That is why the beverage can producer demanded a forecasting system that would help to create an annual production plan. The aim is to analyze the ordering process for the last two years and on that basis to create a forecast system of the possible ordering process for the following year. This is necessary for the introduction of regularity of production, because the frequent transitions of the line to another assortment range or other surface printing requires long downtimes due to the technological setting, thus creating large losses due to inactivity of the production line. As drinking habits of final customers reflect the sale of cans, it was expected that sale data would have a seasonable character; this was proved after a brief analysis of the former data. After that, the appropriate forecasting methods were chosen. The methodology was created to combine multiple forecast results into one to increase the forecast objectivity. Forecasting is performed at three different levels: forecasting of assortment, forecasting of region sale and total forecasting of all orders. In spite of the change in market behavior in 2020, due to the pandemic situation in the first wave of the COVID-19 crisis, the sale of beverage cans is expected to stabilize and return to pre-crisis level as early as 2021. Then the forecasting system will fully meet the company’s requirements.

中文翻译：

---

创建饮料罐生产的年度订单预测——案例研究

本文的重点是创建一个系统，用于预测特定饮料罐制造商的未来订单。问题来自不同客户的罐头不规则订购；不仅是国内公司，还有国外公司。这会导致生产波动，从而导致不规则的运输方式。这就是饮料罐生产商需要一个有助于制定年度生产计划的预测系统的原因。目的是分析过去两年的订购流程，并在此基础上创建下一年可能的订购流程的预测系统。这是引入生产规律所必需的，因为由于技术设置，生产线频繁转换到另一个分类范围或其他表面印刷需要很长时间的停机时间，从而由于生产线的不活动而造成巨大损失。由于最终客户的饮酒习惯反映了罐头的销售情况，预计销售数据具有季节性；在对以前的数据进行简要分析后证明了这一点。之后，选择合适的预测方法。该方法旨在将多个预测结果合二为一，以提高预测的客观性。预测在三个不同的级别执行：分类预测、区域销售预测和所有订单的总预测。尽管 2020 年市场行为发生了变化，但由于第一波 COVID-19 危机的大流行情况，