Optimized hardware product returns for subscription services

Patent metadata

Field	Value
Publication number	US-12136095-B2
Application number	US-202217659904-A
Country	US
Kind code	B2
Filing date	Apr 20, 2022
Priority date	Apr 20, 2022
Publication date	Nov 5, 2024
Grant date	Nov 5, 2024

Abstract

In one aspect, an example methodology implementing the disclosed techniques includes, by a product subscription service, receiving information regarding a hardware asset being returned at an end of a subscription and predicting, using a first machine learning (ML) model, whether the hardware asset has reached EOL. The method also includes, responsive to predicting that the hardware asset has reached EOL, creating, by the product subscription service, a work order to dispatch an eco-partner. The method may further include, by the product subscription service, responsive to predicting that the hardware asset has not reached EOL, predicting, using a second ML model, one or more new subscription orders matching the hardware asset and recommending the one or more matching new subscription orders as possible fits for the hardware asset.

First claim

What is claimed is:

1. A method comprising:
training, at a product subscription service, a first machine learning (ML) model to predict whether a hardware asset has or has not reached end-of-life (EOL) using a training dataset composed of a plurality of training samples, each training sample of the plurality of training samples corresponding to historical recycling and reverse logistics data, each training sample of the plurality of training samples to adjust weights in the first ML model, wherein training the first ML model includes inputting different portions of the training dataset and comparing predictions of EOL with actual target values of the training samples to adjust weights in the first ML model;
receiving, by the product subscription service, information regarding a hardware asset being returned at an end of a subscription;
determining, by the product subscription service, one or more relevant features from the information regarding the hardware asset, the one or more relevant features including hardware performance data and influencing prediction of the EOL;
generating, by the product subscription service, a feature vector including the one or more relevant features;
inputting, by the product subscription service, the feature vector to the first ML model;
predicting, by the product subscription service using the first ML model, whether the hardware asset has reached EOL based on the one or more relevant features; and
responsive to predicting that the hardware asset has reached EOL, creating, by the product subscription service, a work order to dispatch an eco-partner.

2. The method of claim 1 , wherein the first ML model includes a deep neural network (DNN)-based classifier.

3. The method of claim 1 , wherein the work order dispatches the eco-partner to a customer location to pick-up the hardware asset.

4. The method of claim 1 , further comprising, responsive to predicting that the hardware asset has not reached EOL:
predicting, by the product subscription service using a second ML model, one or more subscription demands matching the hardware asset; and
recommending, by the product subscription service, one or more new subscription orders associated with the predicted one or more subscription demands as possible fits for the hardware asset.

5. The method of claim 4 , wherein the second ML model includes a k-nearest neighbor (k-NN) classifier.

6. The method of claim 4 , wherein the one or more subscription demands are predicted using one of Euclidean distance, Manhattan distance, or Minkowsiki distance.

7. The method of claim 4 , further comprising, responsive to selection of one of the recommended one or more new subscription orders, initiating, by the product subscription service, a reverse logistics workflow for the selected one of the recommended one or more new subscription orders.

8. The method of claim 4 , further comprising, responsive to selection of none of the recommended one or more new subscription orders, initiating, by the product subscription service, a workflow for return of the hardware asset.

9. The method of claim 1 wherein the first ML model is a binary classification model.

10. The method of claim 9 wherein the first ML model comprises a rectified linear unit activation function.

11. A system comprising:
one or more non-transitory machine-readable mediums configured to store instructions; and
one or more processors configured to execute the instructions stored on the one or more non-transitory machine-readable mediums, wherein execution of the instructions causes the one or more processors to carry out a process comprising:
training a first machine learning (ML) model to predict whether a hardware asset has or has not reached end-of-life (EOL) using a training dataset composed of a plurality of training samples, each training sample of the plurality of training samples corresponding to historical recycling and reverse logistics data, each training sample of the plurality of training samples to adjust weights in the first ML model, wherein training the first ML model includes inputting different portions of the training dataset and comparing predictions of EOL with actual target values of the training samples to adjust weights in the first ML model;

receiving information regarding a hardware asset being returned at an end of a subscription;
determining one or more relevant features from the information regarding the hardware asset, the one or more relevant features including hardware performance data and influencing prediction of the EOL;
generating, by the product subscription service, a feature vector including the one or more relevant features;
inputting, by the product subscription service, the feature vector to the first ML model;
predicting, using the first ML model, whether the hardware asset has reached EOL based on the one or more relevant features; and
responsive to predicting that the hardware asset has reached EOL, creating a work order to dispatch an eco-partner.

12. The system of claim 11 , wherein the first ML model includes a deep neural network (DNN)-based classifier.

13. The system of claim 11 , wherein the work order dispatches the eco-partner to a customer location to pick-up the hardware asset.

14. The system of claim 11 , wherein the process further comprises, responsive to predicting that the hardware asset has not reached EOL:
predicting, using a second ML model, one or more subscription demands matching the hardware asset; and
recommending one or more new subscription orders associated with the predicted one or more subscription demands as possible fits for the hardware asset.

15. The system of claim 14 , wherein the second ML model includes a k-nearest neighbor (k-NN) classifier.

16. The system of claim 14 , wherein the one or more subscription demands are predicted using one of Euclidean distance, Manhattan distance, or Minkowsiki distance.

17. The system of claim 14 , wherein the process further comprises, responsive to selection of one of the recommended one or more new subscription orders, initiating a reverse logistics workflow for the selected one of the recommended one or more new subscription orders.

18. The system of claim 14 , wherein the process further comprises, responsive to selection of none of the recommended one or more new subscription orders, initiating a workflow for return of the hardware asset.

19. A method comprising:
training, at a product subscription service, a machine learning (ML) model to predict whether a returned hardware asset matches a subscription demand using a training dataset composed of a plurality of training samples, each training sample of the plurality of training samples corresponding to historical subscription demand data, each training sample of the plurality of training samples to adjust weights in the ML model, wherein training the ML model includes inputting different portions of the training dataset and comparing predictions of matches with actual target values of the training samples to adjust weights in the ML model;
receiving, by the product subscription service, information regarding a subscription demand from a new subscription order;
retrieving, by the product subscription service, information regarding a plurality of returned hardware assets, the information including hardware performance data and influencing prediction of an end-of-life (EOL);
generating, by the product subscription service, a feature vector including one or more relevant features from the plurality of returned hardware assets;
i

Assignees

• Dell Products Lp

Inventors

• Mohanty Bijan Kumar
• Kumar Dhilip
• Sahoo Sujit Kumar
• Dinh Hung

Classifications

• G06Q30/014

  Providing recall services for goods or products · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title

• G06N3/09

  Supervised learning · CPC title