Association analysis and determining product locations with machine learning in operation center of e-commerce platforms

The epidemic disease caused by the coronavirus, which first appeared in Wuhan, China in December 2019 and was later declared a pandemic, affected the whole world in a short time in health, economy, business life, social life, education and many other areas. The coronavirus has also accelerated the transformation of the trade that is taking place, in other words, the process of shifting from the physical environment to the virtual environment. E-commerce is one of the sectors that will be most affected in this process. The entry of the pandemic into our lives has increased the interest in the e-commerce sector and contributed greatly to the development of the sector. In this period, the e-commerce sector has grown globally. The interest in the rapidly growing e-commerce sector is increasing day by day. The pandemic has brought some difficulties along with its contribution to the development of the sector. The Pandemic, in the period when everyone was closed to their homes and afterwards, has also led to a change in the shopping habits of many people. This change has caused e-commerce platforms operating in the sector to experience various difficulties in storage operations. Therefore, companies operating in the sector need to increase the importance they attach to storage activities in order to respond to customer needs. There are different processes in the warehouse. The productivity gains to be achieved in these processes have an impact on the performance of warehouse operations. The most important of these processes is the order picking process. The order picking process may differ due to both meeting customer needs and being affected by sectoral development. Order picking is the process of selecting and picking products from a warehouse to fulfill customer orders. Order picking is a very important process in the e-commerce industry as it directly affects customer satisfaction and delivery time. In most companies it is still predominantly done manually by human operators, and this is costly due to significant manual labor. Efficient order picking is critical to ensuring e-commerce success. Customers expect timely and accurate deliveries, and any delay or error in the order picking process can lead to customer dissatisfaction and lost revenue. Additionally, inefficient order picking can lead to higher operational costs, reducing the profitability of the e-commerce business. On the other hand, efficient order picking can lead to increased customer satisfaction and repeat business. It can also increase operational efficiency, reduce costs and increase profitability. While today's technology is developing rapidly, companies want to make maximum use of these technologies in their operations. With the developing technology and the growth of the sectoral volume, big data has started to be produced and stored in companies. The stored data has created the need for companies to produce more meaningful results with the teaching of their machines. These needs have begun to be met by working with these data together with machine learning algorithms. Machine learning is a branch of computer science that allows computers to learn directly without being programmed. Algorithms have been increasingly used to analyze these collected datasets and create decision-making systems to come up with an algorithmic solution to complex problems. This study covers the analysis of customer orders using the Apriori algorithm, one of the machine learning algorithms of customer habits, and the replanning of the warehouse shelf layout with the results obtained. In this study, the sales data of Trendyol, an e-commerce platform operating in Turkey, are discussed. The data includes customer order number, customer order group number, product category, location and quantity ordered. When the data discussed are examined, it consists of a total of 100,004 product sales and 100,000 lines. Among the 12 product categories in total, the most preferred products were TekstilC, TekstilA and TekstilE. The problem discussed in the study and the model applied for its solution are very important in terms of warehouse operations. However, it will be least affected by the changing sector and customer habits and will be able to adapt quickly. In addition, the shortening of customer order preparation times will contribute to increasing customer satisfaction and increasing sales volume. In this study, the data and the problem are introduced. Afterwards, Apriori algorithm was applied to these data in Python programming language. Python is a versatile program used by data analysts and other professionals to perform complex statistical calculations, create data visualizations, create machine learning algorithms, manipulate and analyze data, and complete other data-related tasks. The Apriori algorithm finds the relationship between objects. It refers to the algorithm used to calculate the association rules. It reveals how two or more objects are related to each other. In other words, the apriori algorithm is an association rule that analyzes whether people who buy product A also buy product B. In this algorithm, the association relationship between the product categories of grouped customer orders is analyzed. Then, the algorithm was designed and implemented in Python programming language in order to address the product categories with higher correlation density as a result of the analysis, in closer locations in the warehouse shelf layout. With this algorithm, a location suggestion system has been developed to address the product category. With the proposed method in this study, it was also seen that the warehouse personnel had a positive effect on the hourly product picking productivity. It has been observed that an efficiency of 31.8% was achieved in the hourly product collection performance of the warehouse personnel. Thus, with the location suggested for the product categories in this study, it is aimed to collect customer orders by walking less distance and to have a more efficient warehouse operation for the company. Finally, suggestions are shared in order to take this study further.