We plan to leverage artificial intelligence techniques and external data sources to discover hidden sales patterns and improve SUK sales forecasts, at both local and regional levels. Based on these predictions, we will use mixed integer programming to minimise security stock levels and determine the optimal volume of orders, subject to out-of-stock level constraints.

Our proposed solution includes both a visualisation of our generated insights and a detailed report of the customer data. It will explore a wide range of machine learning methods used for time series predictions, from classical ones (exponential moving average, ARIMA) to more advanced ones (Bayesian methods, deep neural networks such as LSTMs and RNNs). Specifically, we want to leverage the unique expressive power of neural networks to extract hidden patterns from an array of datasets. By combining Bayesian methods with deep neural networks, we aim to model the uncertainty in our predictions, which allows for more acc...
We plan to leverage artificial intelligence techniques and external data sources to discover hidden sales patterns and improve SUK sales forecasts, at both local and regional levels. Based on these predictions, we will use mixed integer programming to minimise security stock levels and determine the optimal volume of orders, subject to out-of-stock level constraints.

Our proposed solution includes both a visualisation of our generated insights and a detailed report of the customer data. It will explore a wide range of machine learning methods used for time series predictions, from classical ones (exponential moving average, ARIMA) to more advanced ones (Bayesian methods, deep neural networks such as LSTMs and RNNs). Specifically, we want to leverage the unique expressive power of neural networks to extract hidden patterns from an array of datasets. By combining Bayesian methods with deep neural networks, we aim to model the uncertainty in our predictions, which allows for more accurate business decisions. These approaches enable us to seamlessly integrate new data sources and create robust solutions, based on a consistent core framework that addresses the problem's full complexity. Besides the provided internal data (historical sales, store locations, etc.), we would like to incorporate additional internal data sources (e.g. historical item prices) and multiple external data sources. The latter, which includes weather patterns, special events, demographics and financial regional statistics, can be obtained from open source data providers. We would also like to incorporate other external data, such as competitor pricing. We believe that there is tremendous potential in quantifying customer behaviour at retail locations (tracking movement and time spent evaluating products) and online.

We believe that we can turn the original stock-security optimisation problem into a profit maximisation one, subject to various internal and external constraints. We aim to estimate the cost relating to stockout and security stocks, and taking into account individual product margin performance. Other factors include logistics costs, the expiry dates for products and intangible factors such as loss of brand prestige due to item stockout.
More information

Investors

Employees

Paul Pop
Admin
Paul Pop CEO I'm a passionate data scientist, looking to use the latest Machine learning algorithms to solve real world problems.

Join the team