View All Reference Kits

Background

The rise of e-commerce combined with the COVID-19 pandemic has permanently changed consumers' expectations around the delivery speed of goods, stretching the retailer supply chain.

One study states that 40 percent of US consumers were willing to wait for two days for orders delivered via fast shipping.¹ Another study found that 69 percent of respondents are less likely to shop with a retailer in the future if their purchase is not delivered within two days of the date promised.²

In addition to delivery speed, the ability to track the movement of a parcel and anticipated delivery window with notification of delivery status is important to customers.³ This increased transparency offers greater customer experience and increases long-term brand loyalty.⁴

On-time delivery forecasting and the rise of transparency of delivery status are opportunities to use AI and machine learning to forecast within the supply chain variability. AI and machine learning for delivery forecasting is highly CPU- and memory-intensive due to the real-time nature of the data and the frequent updates. This reference kit features Intel® software products, which are optimized for performance of machine learning model training and inference. Paired with Intel software products, this reference kit delivers faster order-to-deliver time forecasting.

Solution

In collaboration with Accenture*, Intel developed an order-to-delivery time forecasting reference kit, which may assist your machine learning-based application to provide delivery time forecasting insights. This reference kit includes:
 

• Training data
• An open source, trained model
• Libraries
• User guides
• Intel® AI software components

With a machine learning-based predictive application, retailers can take proactive actions to mitigate delays, costs, and loss of revenue, and then share the delivery details with their customers.

This reference kit builds the order-to-delivery time forecasting model using three algorithms: XGBoost, Random Forests (RF), and Support Vector Machines (SVM). These three analytical models are then fed into an ensembling technique called Voting Model. Voting Model consists of a voting regressor and voting classifier, which then combines the individual predictions to provide a final, consensus prediction to improve the accuracy of the model.

The Intel® AI Analytics Toolkit (AI Kit) gives the data scientist, AI developer and researchers familiar Python* tools and frameworks to accelerate the end-to-end data and analytics pipelines on Intel® architectures. The components are built using AI Kit for low-level compute optimizations. Specifically, the Intel® Extension for Scikit-learn* and daal4py are used to accelerate the individual build components as well as the ensemble model build.

At a Glance

• Industry: Retail, e-commerce, supply chain management
• Task: Predict wait time for shipment delivery
• Dataset: Anonymized dataset from e-commerce company
• Type of Learning: Supervised machine learning
• Models: Voting Model:
  • Voting regression for predicting wait time
  • Voting classifier for predicting likelihood of delay
• Output: Wait time and likelihood of delay are the predicted variables.
• Intel® AI Software Products:
  • AI Kit
  • Intel® Optimization for XGBoost*
  • Intel® Extension for Scikit-learn*

Technology

Optimized Intel® AI Software Products for Better Performance.

AI Kit—Achieve end-to-end performance for AI workloads.

Intel Optimization for XGBoost

Intel has been directly upstreaming many optimizations to provide improved performance on Intel CPUs. This well-known, machine learning package for gradient-boosted decision trees now includes seamless, drop-in acceleration for Intel® architectures to significantly speed up model training and improve accuracy for better predictions.

Performance was tested on Microsoft Azure* Standard_D8s_V5 instances powered by 3rd generation Intel® Xeon® processors to optimize the solution.

Benefits

For the enterprise developer, the AI Kit is optimized for machine learning training and inference while unlocking additional compute capacity. This kit offers an machine learning-based algorithm for providing delivery time forecasting insights. With this machine learning-based predictive reference kit, retailers can take proactive actions to help mitigate delays, costs, and loss of revenue, and then share the delivery journey details with their customers.

Download Kit

Related Reference Kits

Personalize Retail Experiences with Enhanced Customer Segmentation

Efficiently Automate Retail Purchase Prediction

Additional Resources

AI Kit

Intel® AI Software Portfolio

Developer Resources from Intel and Accenture

References

1. Fast Online Order Delivery According to US Consumers 2019. Stephanie Chevalier, July 27, 2022, https://www.statista.com/statistics/561768/fast-online-order-delivery-us-consumers/
2. The Impact of Late and Inaccurate Deliveries on Customer Loyalty, Alison Howen, posted December 7, 2014, Website Magazine, https://www.websitemagazine.com/blog/the-impact-of-late-and-inaccurate-deliveries-on-customer-loyalty
3. What Do Customers Actually Want When It Comes to Delivery? Will Gillingham, IMRG: The UK Ecommerce Association, 2022. https://www.imrg.org/blog/what-do-customers-actually-want-when-it-comes-to-delivery/
4. Consumers Want Transparency. Here’s How to Give It to Them. The Light Digital Blog, 2022. https://www.labelinsight.com/Transparency-ROI-Study