ACCELERATE YOUR JOURNEY AS AN AI-DRIVEN COMPANY

AI-POWERED FORECASTING TOOL

BRIEF

Who doesn’t want to predict the future? At Mediaan Conclusion, we built a data forecasting tool that allows you to make smarter decisions with AI-powered predictions. The purpose of this tool is to help our clients kick-start or scale up their data science forecasting projects and become more data-driven organizations. It reduces time and effort to implement forecasting models by using data generation models, proven out-of-the-box forecasting algorithms, and model management capabilities.

THE CONCERN

While forecasting using AI allows for smarter and more profitable decision-making, many businesses are still reluctant to take the first leap towards implementing AI solutions. Another case is that they have already started with a proof-of-concept but have no idea what the next steps would be. The main factors are often a lack of resources such as time, expertise, or qualified data.

THE CONCEPT

“How can Mediaan Conclusion create a solution that helps our clients to complete a full cycle of AI solutions and accelerate their journey to becoming AI-driven companies?” The idea was to build a user-friendly forecasting tool that would work whether or not the company that uses it has data and expertise. Hence, we built the tool with the four steps of the data science pipeline in mind; data handling, model development & management, IT integration, and presentation of results.

DATA SCIENCE PIPELINE IN DETAIL

Built based on the four steps of the data science pipeline, the forecasting tool offers the following benefits at your disposal:

Data handling. The tool allows data to be imported, connected, or created. Even without any data, a company can start forecasting by generating high-quality synthetic data from business parameters. These parameters can simulate signals from multiple industries, such as sales, energy, IoT, weather, customer behavior, and many more.
Model development & management. The core of AI is the algorithm at disposal. Therefore, the tool offers many out-of-the-box algorithms that can be trained and validated. With the help of MLFlow, the best forecasting model can be selected, and become reproducible for production environments.
IT integration. The tool was built to be either a stand-alone product or to be integrated into the IT landscape of any company. Built from open-source technologies, making it affordable.
Presentation of results. Historical data and forecasts are visually shown for the purpose of model improvement or real business decision-making. Another option is to integrate forecasting results into existing systems, databases, or BI tools.

TECHNOLOGY

The tool was created by the Data & Intelligence team of Mediaan Conclusion, using the following methods and technologies:

Python
Streamlit
Timesynth
scikit-learn
TensorFlow
Darts
FBProphet
Exponential Smoothing
ARIMA/SARIMA

Linear Regression
Random Forest Regression
N Beats
XGB Regression
Transformers
Temporal Convolutional Networks
Long short-term memory (LSTM)

FUTURE PLANS

Future plans include the expansion of the number of techniques that can be used for forecasting. There are also plans to introduce templates that make it easier to generate domain-specific data without much hassle. Furthermore, there are plans of introducing models that can forecast based on hierarchically constrained-sets of forecasting data.

RESULTS

Our forecasting tool supports companies through their journey of completing a full cycle of an AI solution, and helps accelerate their journey in becoming data-driven companies. We made sure that the tool does all the heavy lifting of data modelling, with optimization engines and state-of-the-art machine learning running in the background. A tool that provides intelligent forecasting and improves your process, thus minimizing risk, reducing costs, and saving time and resources.

WANT TO SEE MORE?