solution.yaml Configuration¶

This document aims to explain the parameters used in the conf/solution.yaml file and their effects.

Supported Recipes and Corresponding API¶

Recipe	Description	Recommend API	AWS Reference
user_personalization	Provides personalized recommendations for users based on their behavior	`recommend-items-to-user`	User Personalization
similar_items	Finds items similar to those that users have interacted with	`recommend-items-to-item`	Similar Items
personalized_ranking	Re-ranks a list of items for each user based on their preferences	`recommend-items-to-user`	Personalized Ranking

Personalize Recipes Configuration¶

user_personalization¶

Configuration for the user personalization recipe.

recipes:
  user_personalization:
    target_event_type: "your_target_event"
    event_value_threshold: "0"
    parameters:
      - name: "hidden_dimension"
        value: 128
    excluded_dataset_columns:
      - name: "Item"
        value:
          - ""
          - ""
      - name: "User"
        value:
          - ""
      - name: "Interaction"
        value:
          - ""

similar_items¶

Configuration for the similar items recipe.

recipes:
  similar_items:
    target_event_type: "your_target_event"
    event_value_threshold: "0"
    parameters:
      - name: "popularity_discount_factor"
        value: 0.5
    excluded_dataset_columns:
      - name: "Item"
        value:
          - ""
          - ""
      - name: "User"
        value:
          - ""
      - name: "Interaction"
        value:
          - ""

personalized_ranking¶

Configuration for the personalized ranking recipe.

recipes:
  personalized_ranking:
    target_event_type: "your_target_event"
    event_value_threshold: "0"
    parameters:
      - name: "bptt"
        value: 8
    excluded_dataset_columns:
      - name: "Item"
        value:
          - ""
          - ""
      - name: "User"
        value:
          - ""
      - name: "Interaction"
        value:
          - ""

Common Parameters¶

target_event_type¶

Specifies the event type that triggers the recommendation process.

Note

This should match the event type defined in the dataset schema.

For example, if your event dataset is as follows, then this field should be set to rating:

USER_ID,ITEM_ID,EVENT_TYPE,EVENT_VALUE,TIMESTAMP
1,1193,rating,5,978300760
1,661,rating,3,978302109
1,914,rating,3,978301968

event_value_threshold¶

Sets the threshold value for considering an event in the recommendation process. During training, only event values greater than the threshold will be used.

parameters¶

Defines a list of parameters to customize the recipe. Each recipe supports different parameters. Please refer to the corresponding AWS recipe documentation to see the available parameters.

excluded_dataset_columns¶

Specifies the columns to exclude from training. Each key is a dataset type, and each value is a list of columns. Exclude columns to control what data Amazon Personalize uses to generate recommendations.