EXCEPTION PREDICTION TOOL FOR SUPPLY CHAIN | Unicsoft

Exception Prediction Tool for Supply Chain

EXCEPTION PREDICTION TOOL FOR SUPPLY CHAIN

Machine Learning for automation

Challenge

Modern global supply chains are becoming vast and increasingly dynamic ecosystems. Supply chain optimization becomes the goal for almost all companies to gain competitive advantages on the market.

The client is a leading provider of supply chain consulting, software and fourth-party logistics services. Aiming to bring more business value to their clients, they strive to optimize logistics execution, using Machine Learning for automation of exception prediction and data processing from different suppliers and thus facilitating the process of decision making. There is a strategic R&D department inside the client’s company, which works with Unicsoft on improving their current solution by making it more efficient, accurate and sophisticated in terms of predicting errors in logistics.

Our clients’ R&D department works on the Proof of Concepts. And Unicsoft helps to transform PoCs with the potential of developing into a solution as a feature of the existing product. It helps to improve the product via optimizing of logistics operations.

Solution

As the time to market was of the highest priority we have established very tight customer-Unicsoft cooperation with bi-weekly sync-ups and direct communication between technical teams to complete the project on agreed delivery date.

We have deeply reworked the Machine Learning model for logistics incidents prediction and supplier-related incidents. The tool takes the input from the past for training and then trained model is used to predict exceptions next week. It focuses on forecasting a group of exceptions. The tool predicts if the consignor provides a higher or lower volume than advised or the consignor fails to ship at all. The tool is used every Thursday and does a prediction for the upcoming week. It was programmed in R.

Result and Future Plans

Unicsoft team considerably improved the machine learning model as well as other components of the solution and as a result, the client received:

  • Higher precision for the forecasts.
  • The user interface that was recreated from scratch to make the tool easier to use with a modern interface.
  • Opportunity to generate reports in a number of formats and with needed periodicity.

Being satisfied with the project success the client asked us to complement the solution with analysis and prediction of the carrier-related incidents (last mile delivery) in the nearest future. Also, Unicsoft team will be responsible for the research and development of the product’s other concepts regarding the precision for the supply chain forecasts.

TECHNOLOGY & TOOLS: RR ShinyMachine LearningPredictive Analytics