PhD Research Seminar: Maximizing Retailer Profit Using an Electronic Price-Tag System, Deep Generative Models for Anomaly Detection
First talk: Maximizing Retailer Profit Using an Electronic Price-Tag System
Speaker: Kirill Kanishchev, second-year PhD student, Department of Business Informatics
Each product in a store has a certain interval of potential prices limited by the maximum and minimum acceptable selling price due to the general pricing policy, expiration dates, warehouse stocks, and the economic situation. Depending on the current values of various factors, the price on the product is set within the entire permissible interval. However, the demand functions of different customers for the same product differ from each other and are determined by their utility functions and income levels. In general terms, we can say that each individual has one's own function of the price of a particular product, which depends on a large number of factors. The current state of affairs in retail chains that use standard paper price-tags unifies all customers by offering them the same terms of purchase for the same product. As a result, the pricing policy and financial indicators are usually sub-optimal, and there is no opportunity to apply the principles of price discrimination. This causes a delayed reaction to changing market conditions and loss of profits. At the same time, thanks to the use of electronic price-tags, the price of each product item can change many times during the day, depending on the regularities established by the CRM system between the final demand for a particular product and such parameters as, for example, the consumer population, day of the week, time, weather and special offers. At the same time, the electronic price tags themselves are only a tool for influencing the demand and financial performance of the retailer, while the key success factor is the principle of dynamic pricing. Moreover, it is the bottom line pricing that is the factor that directly affects profitability. The target “correct” price should be such as to bring the maximum profit to the retailer in a given period of time for a given category of buyers in a given segment of goods. Thus, to maximize the retailer's profits, a complex solution is required that will be responsible for finding the mathematically correct price for each product in each store in the chain at a specific time interval, taking into account the elasticity of demand, the current competitive environment, traffic, and many other related factors. Accordingly, the purpose of the work is to research how it is necessary to organize the process of using electronic price-tags to maximize the profit of a retailer; how the use of a CRM system, NFC technology, and a loyalty program can help in this process; how it is necessary to algorithmize the dynamic price updates in order not only to maximize the total profit, but also to minimize the total costs for each section of analysis.
Second talk: Deep Generative Models for Anomaly Detection
Speaker: Artem Ryzhikov, seond-year PhD student, Faculty of Computer Science
Anomaly detection for complex data is a challenging task from the perspective of machine learning. In this work, we consider cases with a small number of anomalies given in the training dataset, while significant statistics for the normal class are available. For such scenarios, conventional supervised methods might suffer from the class imbalance, while unsupervised methods tend to ignore difficult anomalous examples. We extend the idea of the supervised classification approach for class-imbalanced datasets by exploiting generative models.