Personalizeのドメイン最適化レコメンデーションを使ってみた

import boto3
from boto3.session import Session
import json
import pathlib
import pandas as pd
import numpy as np

df_ratings = pd.read_csv("ml-latest-small/ratings.csv")

# columnsの書き換え
df_ratings = df_ratings.rename(columns={'userId': 'USER_ID', 'movieId': 'ITEM_ID', 'timestamp': 'TIMESTAMP'})

# ratingを削除
df_ratings = df_ratings.drop('rating', axis=1)

# EVENT_TYPEを生成
rns = np.random.RandomState(seed=777)
list_event_type = rns.choice(['Watch', 'Click'], size=len(df_ratings), replace=True, p=[0.8,0.2])
df_ratings['EVENT_TYPE'] = list_event_type

df_ratings.to_csv("ml-latest-small/ratings_mod.csv", index=False)

bucket_name = 'Your S3 bucket name'
import_uri = 'Your S3 file uri' # s3://bucket_name/file_path'

region_name = 'ap-northeast-1'
prefix = 'trial-20220222'

dataset_group_name          = f'{prefix}-dataset-group'
schema_interaction_name     = f'{prefix}-schema-interaction'
dataset_interaction_name    = f'{prefix}-dataset-interaction'
import_job_interaction_name = f'{prefix}-import-job-interaction'
recommender_name            = f'{prefix}-recommender-vod-most-popular'
iam_role_name               = f'{prefix}-personalize-exection-role'
iam_custom_policy_name      = f'{prefix}-personalize-execution-policy'

client_s3                  = boto3.client('s3', region_name=region_name)
client_personalize         = boto3.client('personalize', region_name=region_name)
client_personalize_runtime = boto3.client('personalize-runtime', region_name=region_name)
client_iam                 = boto3.client('iam', region_name=region_name)

# create_bucket
location = {'LocationConstraint': region_name}
client_s3.create_bucket(Bucket=bucket_name, CreateBucketConfiguration=location)

# put_public_access_block
client_s3.put_public_access_block(
    Bucket=bucket_name,
    PublicAccessBlockConfiguration={
        'BlockPublicAcls': True,
        'IgnorePublicAcls': True,
        'BlockPublicPolicy': True,
        'RestrictPublicBuckets': True,
    },
)

# create bucket_policy
bucket_policy = {
    "Version": "2012-10-17",
    "Id": "PersonalizeS3BucketAccessPolicy",
    "Statement": [
        {
            "Sid": "PersonalizeS3BucketAccessPolicy",
            "Effect": "Allow",
            "Principal": {
                "Service": "personalize.amazonaws.com"
            },
            "Action": [
                "s3:GetObject",
                "s3:ListBucket"
            ],
            "Resource": [
                f"arn:aws:s3:::{bucket_name}",
                f"arn:aws:s3:::{bucket_name}/*"
            ]
        }
    ]
}
bucket_policy = json.dumps(bucket_policy)

# put_bucket_policy
client_s3.put_bucket_policy(
    Bucket=bucket_name,
    Policy=bucket_policy,
)

interactions_csv = pathlib.Path('ml-latest-small/ratings_mod.csv')
response = client_s3.upload_file(
    str(interactions_csv),
    bucket_name, 
    str(interactions_csv),
)

response_create_dataset_group = client_personalize.create_dataset_group(
    name=dataset_group_name,
    domain='VIDEO_ON_DEMAND'
)

schema_interaction = {
    "type": "record",
    "name": "Interactions",
    "namespace": "com.amazonaws.personalize.schema",
    "fields": [
        {
            "name": "USER_ID",
            "type": "string"
        },
        {
            "name": "ITEM_ID",
            "type": "string"
        },
        {
            "name": "TIMESTAMP",
            "type": "long"
        },
        {
            "name": "EVENT_TYPE",
            "type": "string"
        },
    ],
    "version": "1.0"
}

schema_interaction = json.dumps(schema_interaction)

response_create_schema = client_personalize.create_schema(
    name=schema_interaction_name,
    schema=schema_interaction,
    domain='VIDEO_ON_DEMAND',
)

response_create_dataset = client_personalize.create_dataset(
    name=dataset_interaction_name,
    schemaArn=response_create_schema['schemaArn'],
    datasetGroupArn=response_create_dataset_group['datasetGroupArn'],
    datasetType='Interactions'
)

iam_custom_policy_document = {
    "Version": "2012-10-17",
    "Statement": [
        {
            "Action": [
                "s3:ListBucket"
            ],
            "Effect": "Allow",
            "Resource": [
                f"arn:aws:s3:::{bucket_name}"
            ]
        },
        {
            "Action": [
                "s3:GetObject",
                "s3:PutObject"
            ],
            "Effect": "Allow",
            "Resource": [
                f"arn:aws:s3:::{bucket_name}/*"
            ]
        }
    ]
}

response_create_policy = client_iam.create_policy(
    PolicyName=iam_custom_policy_name,
    PolicyDocument=json.dumps(iam_custom_policy_document),
)

assume_role_policy_document = {
    "Version": "2012-10-17",
    "Statement": [
        {
          "Effect": "Allow",
          "Principal": {
            "Service": "personalize.amazonaws.com"
          },
          "Action": "sts:AssumeRole"
        }
    ]
}

create_role_response = client_iam.create_role(
    RoleName = iam_role_name,
    AssumeRolePolicyDocument = json.dumps(assume_role_policy_document)
)

client_iam.attach_role_policy(
    RoleName = iam_role_name,
    PolicyArn = "arn:aws:iam::aws:policy/AmazonS3ReadOnlyAccess"
)

client_iam.attach_role_policy(
    RoleName = iam_role_name,
    PolicyArn = response_create_policy['Policy']['Arn']
)

response_get_role = client_iam.get_role(RoleName=iam_role_name)

response_create_dataset_import_job = client_personalize.create_dataset_import_job(
    jobName=import_job_interaction_name,
    datasetArn=response_create_dataset['datasetArn'],
    dataSource={
        'dataLocation': import_uri
    },
    roleArn=response_get_role['Role']['Arn']
)

response_create_recommender = client_personalize.create_recommender(
    name=recommender_name,
    datasetGroupArn=response_create_dataset_group['datasetGroupArn'],
    recipeArn='arn:aws:personalize:::recipe/aws-vod-most-popular',
)

response_get_recommendations = client_personalize_runtime.get_recommendations(
    userId="1",
    numResults=5,
    recommenderArn=response_create_recommender['recommenderArn'],
)

{
  "ResponseMetadata": {
    "RequestId": "9e8ec5f8-4a68-46d2-849f-14432877395a",
    "HTTPStatusCode": 200,
    "HTTPHeaders": {
      "content-type": "application/json",
      "date": "Mon, 21 Feb 2022 07:22:11 GMT",
      "x-amzn-requestid": "9e8ec5f8-4a68-46d2-849f-14432877395a",
      "content-length": "322",
      "connection": "keep-alive"
    },
    "RetryAttempts": 0
  },
  "itemList": [
    {
      "itemId": "356"
    },
    {
      "itemId": "296"
    },
    {
      "itemId": "318"
    },
    {
      "itemId": "593"
    },
    {
      "itemId": "2571"
    }
  ],
  "recommendationId": "RID-b0c39904-e612-4158-92e3-a1ca3661d508"
}

response_delete_recommender = client_personalize.delete_recommender(
    recommenderArn=response_create_recommender['recommenderArn']
)