Get Recommendations Based on Users' Browsing History

You will use Swagger UI, via any web browser, to call the Personalized Recommendation service APIs. Swagger UI allows developers to effortlessly interact and try out every single operation an API exposes for easy consumption. For more information, see Swagger UI.

In the service key you created for Personalized Recommendation in the previous tutorial: Use the Free Tier Service Plan to Set Up Account for Personalized Recommendation and Get Service Key, you should find (outside the uaa section of the service key) an entry called url (as highlighted in the image below).

1. To access the Personalized Recommendation Swagger UI, add /doc to the url value, paste it into any web browser and press Enter.

   

2. To be able to use the Swagger UI endpoints, you need to authorize yourself. In the top right corner, click Authorize.

   

3. Get the access_token value created in the previous tutorial: Get OAuth Access Token for Personalized Recommendation via Web Browser, then add Bearer in front of it, and enter in the Value field.

   Bearer <access_token>
   

   

4. Click Authorize and then click Close.

   

Use the POST /standard/rs/v1/tenants/{tenant}/jobs/file-upload endpoint to upload data and trigger the machine learning model training job.

1. Click the endpoint name to expand it.

2. Click Try it out.

   

3. Enter tenant name. It is required to enter a name of your choice for tenants. You can use, for example, pr_tutorial.

4. Click on data to download the training data ZIP file. Click Choose File to upload the training data.

5. Enter site name. Same as tenant name, but if you leave the field blank, it will be named automatically as default. You can use, for example, pr_tutorial_site.

   

6. Set serve_model to true for automatic deployment of real-time model serving instance. To trigger the machine learning training, click Execute.

   

   You should receive a response like the following:

   

Use the GET /standard/rs/v1/tenants/{tenant}/jobs/latest endpoint to check the status of the ongoing training job.

1. Click the endpoint name to expand it.

2. Click Try it out.

3. Enter tenant name (pr_tutorial) and site name (pr_tutorial_site). Click Execute.

   

   You should receive a response like the following:

   

   After a few minutes, the training job status should change to SUCCEEDED.

   

Use the GET /standard/rs/v1/tenants/{tenant}/servings endpoint to check the deployment status for model serving.

1. Click the endpoint name to expand it.

2. Click Try it out.

3. Enter site name (pr_tutorial_site) and tenant name (pr_tutorial). Click Execute.

   

   You should receive a response like the following:

   

   After a few minutes, the serving job status should change from PENDING to SUCCEEDED.

   

There are four different choices of inference calls as you can see in Inference Options. In this step, we will use Next-Item Recommendations.

Use the POST /standard/rs/v1/tenants/{tenant}/recommendations/next-items endpoint to get next-item recommendations.

1. Click the endpoint name to expand it.

2. Click Try it out.

3. In the payload, you can enter the following (that means the user has recently clicked an item with ID “2858”):

   JSON

   Copy

   {
      "items_ls":[
         "2858"
      ]
   }
   

4. Enter site name (pr_tutorial_site) and tenant name (pr_tutorial). Click Execute.

   

   You should receive a response like the following:

   

The response shows a sequence of recommendations based on the user’s past clickstreams. Since we included in the payload the item “2858” as our past clickstream, the model is able to recommend 10 other items related to the item “2858”.

The confidence score represents how certain the model is about its prediction. The higher the score the more confident the model is that the prediction is actually correct. If the score is close to 1, the model is very certain.

For this tutorial, we’re using a small dataset that results in a low model quality and that’s why the scores are also low. A bigger and more complex dataset would take much more time to be trained but would provide better scores.

There are four different choices of inference calls as you can see in Inference Options. In this step, we will use Similar-Item Recommendations.

Use the POST /standard/rs/v1/tenants/{tenant}/recommendations/similar-items endpoint to get similar-item recommendations.

1. Click the endpoint name to expand it.

2. Click Try it out.

3. In the payload, you can enter the following (that means you want to get similar-item recommendations for item “1196”):

   JSON

   Copy

   {
      "item":"1196",
      "k":3,
      "explain":true
   }
   

4. Enter site name (pr_tutorial_site) and tenant name (pr_tutorial). Click Execute.

   

   You should receive a response like the following:

   

The response shows 3 similar-item recommendation results ("k":3) based on the item you selected ("item":"1196"). As ML explainability is enabled ("explain":true), in the response, you receive quantifiable insights into how the recommendations engine computes the recommendation results.

There are four different choices of inference calls as you can see in Inference Options. In this step, we will use Smart-search results.

Use the POST /standard/rs/v1/tenants/{tenant}/recommendations/smart-search endpoint to get smart-search results.

1. Click the endpoint name to expand it.

2. Click Try it out.

3. In the payload, you can enter the following (that means the recommendations are personalized by taking into account the user ID ("user":"user_id1") and the current clickstream ("items_ls":["item_id1","item_id2"])):

   JSON

   Copy

   {
      "string_queries":"Lord of the Rings",
      "user":"user_id1",
      "items_ls":[
         "item_id1",
         "item_id2"
      ],
      "explain":true,
      "k":3,
      "attribute_queries":{
         "categoricalFeatures":{
            "tags":{
               "values":[
                  "trilogy",
                  "imdb top 250"
               ],
               "clauses":[
                  "must",
                  "should"
               ]
            }
         }
      }
   }
   

4. Enter site name (pr_tutorial_site) and tenant name (pr_tutorial). Click Execute.

   

   You should receive a response like the following:

   

The response shows, based on a list of free-form text input (string_queries), and/or a list of attribute values (attribute_queries), a number of recommendation results ("recommendations":["1210","1196","1269"]) along with their scores ("scores": [0.3933, 0.3371, 0.6671]). In this example, the item “1210” has a score of 0.3933, the item “1196” has a score of 0.3371, and the item “1269” has a score of 0.6671. When you enable ML explainability, you can better understand the “scores”. For “item_attribute_contribution”, the position of the nested list corresponds to the position of the item in the “recommendations” list. The first nested list [0.0, 0.6691,-0.3309] is the “item_attribute_contribution” for the first item in “recommendations” that is “1210”. And the second nested list [0.0, 0.5117, -0.4883], is the “item_attribute_contribution” for the second item in “recommendations” that is “1196”.

There are four different choices of inference calls as you can see in Inference Options. In this step, we will use User-affinity Recommendations.

Use the POST /standard/rs/v1/tenants/{tenant}/recommendations/user-affinity endpoint to get user-affinity recommendations.

1. Click the endpoint name to expand it.

2. Click Try it out.

3. In the payload, you can enter the following (that means the user has recently clicked the items with IDs “364”, “1064”, and “3108”):

   JSON

   Copy

   {
      "items_ls":[
         "364",
         "1064",
         "3108"
      ],
      "features":[
         "categories",
         "tags"
      ],
      "k":3
   }
   

4. Enter site name (pr_tutorial_site) and tenant name (pr_tutorial). Click Execute.

   

   You should receive a response like the following:

   

The response shows, based on user past item interactions ("items_ls":["364","1064","3108"]) from the request payload, a number of recommended category affinities ("recommendations":{"categories":["Crime","Thriller","Animation"]) along with their scores ("scores":{"categories":[0.2279,0.1805,0.1081]). In this example, the category “Crime” has an affinity score of 0.2279, the category “Thriller” has a score of 0.1805, and the category “Animation” has a score of 0.1081.

Now that you’ve learned the whole process on how to consume the Personalized Recommendation service, it’s time to clean up.

When you trigger the DELETE request, a workflow to delete the model starts. The metering for model serving then stops. Deleting an unused model serving will help free tier users to save inference quota for future usage, for users of the standard service plan, it’ll help to save model serving cost.

Use the DELETE /standard/rs/v1/tenants/{tenant}/servings endpoint to delete the model serving instance of the tenant (site).

1. Click the endpoint name to expand it.

2. Click Try it out.

3. Enter tenant name (pr_tutorial) and site name (pr_tutorial_site). Click Execute.

   

   You should receive a response like the following:

   

Use the DELETE /standard/rs/v1/tenants/{tenant} endpoint to delete all files for any sites that belong to a subaccount and tenant.

1. Click the endpoint name to expand it.

2. Click Try it out.

3. Enter tenant name (pr_tutorial) and site name (pr_tutorial_site).

4. Set all_sites to true and click Execute.

   

   You should receive a response like the following:

   

You have now successfully used the Personalized Recommendation service to get recommendations based on users’ browsing history.