Analytics and Forecasting Based on AI

May 22, 2022

The client needed a solution for predicting the fat and protein content in raw milk - key parameters determining its quality. The client required the ability to make accurate forecasts both in the short term (a weekly forecast for the upcoming month) and in the medium term (a monthly forecast for the next year). Previously, statistical methods were used for these purposes; however, over time, the company realized the need to automate the process and improve forecasting accuracy.


Softline Digital proposed developing a service based on the Ditwinx platform that allows the client to build forecasts interactively. This service was created for the rapid development, deployment, and operation of artificial intelligence-based solutions, assisting clients in optimizing technological processes.

Considering the client's requirements, the solution developed on the Ditwinx platform included the ability to interpret forecast results and assess the impact of various factors. This allows the client to understand which specific factors influence forecasted parameters, such as the fat and protein content in milk. Such interpretation of results can be valuable for making informed decisions and optimizing production processes.

In addition, the service provides the opportunity for a retrospective assessment of the accuracy of the forecasts obtained. This allows the client to analyze and compare forecasts with the actual values of parameters and evaluate the degree of forecast accuracy. Such functionality enables the client to monitor the quality of forecasting and adjust models as needed.

Furthermore, the service includes the capability to build forecasts with a breakdown by milk suppliers. This means that the client can analyze and forecast milk parameters for each supplier separately. This detailed forecasting can be useful for managing milk supplies, analyzing the quality of individual suppliers, and making decisions based on this information.

The development of this service was carried out by experts from Softline Digital, who have experience in developing artificial intelligence-based solutions. They provided the client not only with a functional service based on the Ditwinx platform but also with full support at all stages of the project, including requirements analysis, development, implementation, and user training.

"We decided to develop the service on the Ditwinx platform for several reasons. If we were to use any BI system, it would only provide data visualization, leading to the need to integrate it with other solutions. This would increase the project implementation time and cost. Ditwinx , on the other hand, allows for the creation of complete services within a single ecosystem. The platform includes built-in tools for data analysis and incorporates most machine learning algorithms. In Ditwinx , the customer can not only make forecasts but also adapt models and test their own hypotheses. This is possible thanks to the No-Code core, which allows for self-modification of the system without involving a Data Science team," explains Vadim Sedelnikov, Data Science Architect at Softline Digital.


Specialists at Softline Digital developed a highly accurate and robust forecasting model for key milk parameters and integrated it into the forecasting service with user interfaces tailored to the client's requirements. The increase in forecasting accuracy was achieved by creating a multi-criteria model using machine learning methods and predictor impact analysis.

In addition to the primary task of forecasting the fat and protein content in milk, the Softline Digital team addressed additional sub-tasks. They implemented a function for monthly forecasting of milk fat and protein content with horizons of 8 weeks ahead (weekly) and 18 months ahead (monthly). The ability to build models for a selected date in the past and compare model results with actual data was also added. All of this allows the client to obtain more flexible and detailed information about the forecasted milk parameters.

The client appreciated the effectiveness of the new solution and decided to expand the project, entrusting Softline Digital with the development of a similar service for forecasting market prices for raw milk. This reflects the trust and satisfaction of the client with the results of the Softline Digital team.

The service developed by Softline Digital provides dairy producers with accurate forecasts of qualitative parameters and market prices for raw milk in real-time. This allows the company to efficiently plan and optimize technological processes, as well as make informed decisions in the supply chain.

The services implemented on the Ditwinx platform are seamlessly integrated, providing convenience in usage and integration. Thanks to the developed models and capabilities of the service, the forecasting error was reduced by 30% compared to the client's previous methods.

The service automatically collects data from various sources, processes them, trains models, and builds forecasts in automatic mode. The company can also use the service for independent data analysis by uploading, analyzing, establishing mutual correlations, and building models. This gives employees the opportunity to conduct their own research and test hypotheses.

The solution also provides the ability to import necessary information for analysis and forecasting, as well as export forecast results. It supports versioning of forecasts, allowing tracking of changes and corrections made.

The results of analytics and forecasts are visualized on dashboards, presenting information in a clear and convenient format for the client. This helps the company visually track and analyze data, make informed decisions, and plan further actions.

Overall, the service developed by Softline Digital provides the client with all the necessary tools and functionality for accurate forecasting and analysis of data on qualitative parameters and market prices for raw milk, contributing to the improvement of production processes and informed business decision-making.

Vadim Sedelnikov
Vadim Sedelnikov
Machine Learning Architect at Softline Digital