Enterprises today that have a major dependency on fluctuating economy. One major driver to this fluctuation is the cost of raw materials or commodities. The main objective of global economies is reducing the cost of procuring these commodities. This helps driving margin and profits like wise. The ability to know when to procure these essential raw materials empower organizations to be leader in their businesses.
The goal to commodity price prediction is to make better-informed decisions and to manage costs associated with volatility of the commodity prices. This enables future forecasting prices reliably and accurately. Adding to these benefits these forecasts helps in aiding marketing and speculative strategies.
Several factors influence and impact Commodity Prices, some key factors have been identified by experts as
- Extraction,
- Demand & Supply
- Global Pricing
Commodity prices change frequently, this adds to the challenge in predicting with accuracy.Procurement teams and data analysts track these price fluctuations and make purchase decisions. The most common approach used is targeting the lowest price within a month or any specific period of time. As larger organizations procure in large volumes amounting to millions of dollars, even a single dollar saved per unit adds up to the benefits they will enjoy.
To add value to forward thinking enterprises, our team at VASPP analyzed this ever occurring challenge existent in several industries and created a solution. Using a best fitting machine learning model with help of SAP Data Intelligence, SAP Data Warehousing Cloud and SAP Analytics Cloud this solution enables organizations in approaching this price challenge.
The VASPP solution focuses on the three key aspects:
Data Collection
Data resides in disparate sources like SAP, 3rd party data warehouses, unstructured data like csv/xls/txt, APIs, etc. Using SAP Data Intelligence, the data collection and harmonization are automated.
Accurate Prediction
Price prediction is the task in hand and using ML models in SAP Data Intelligence, our solution helps running several best of bread ML algorithms and promotes the best suited model forward with the least error in prediction. These models are deployed with ease and runs with a very low maintenance effort.
Decision Making
The predictions are pushed into SAP DWC which allows historic and predictive data comparison and simulations. SAP Analytics Cloud helps compare and visualize the predicted prices. Procurement teams can run comparisons and take informed decisions best suited for their cost saving goals.
Solution Overview
The key aspects to the solution follows 3 core steps
Solution Architecture:
The solution uses SAP DWC as a data warehouse where all the predictions and historical data from different sources is collected for analysis.
SAP DI plays the pivotal role, where it extracts the data from sources into DWC followed by data processing. Post data processing, various machine learning models are tested and the model with least error is deployed. The predicted results are stored in DWC models.
The collected data is not in usable format for ML models. Once data is collected, in DI various cleaning and data processing tasks are initiated to fix issues like missing data, outliers, etc. and necessary imputations are made. Pipelines are then created for training and prediction.
In a training pipeline, the data processing is done. After data processing, various ML models are evaluated and best suited model is selected. The next step is hyper-parameter tuning of the selected model to further improve the accuracy and save the model in Artifact producer.
Sample pipelines have been created using the selected framework in SAP Data Intelligence. The pipelines, models and the framework itself can be used as a template. This template has options that are easy to adopt within various types of commodity prediction scenarios.
For more details and technical prerequisites please reach us at analytics@vaspp.com.