Sales analysis applications with machine learning methods

In the retail sector, which is one of the sectors where user habits change most frequently in the world and in our country, it has become difficult for companies to maintain their sales-inventory balances, especially on fast moving consumer goods, without entering economically more advantageous and high stock loads and without adversely affecting product availability. Statistical methods are one of the leading solutions in the construction of this sales-stock balance. In our problem, two-year sales data for cleaning products, paper products, non-food products and personal care product groups of a business that has 171 large-scale stores in Turkey and sells fast moving consumer goods in different categories are obtained from the company in time series and machine learning is used with different artificial intelligence. Sales estimation study of the mentioned product groups was carried out using statistical methods. For the Arima, Sarimax, Triple Exponential Smoothing and Prophet methods used, the estimation results of each category were compared and it was observed that different methods gave optimal results for different categories. The study is presented as an example for the company and as a suggestion that applications can be developed in other categories and regional sales.