Help Center> RES> API Reference> API> Online Service> Creating an Online Service

Creating an Online Service

Function

This API is used to create online service metadata. After the metadata is created, you can manually publish the service.

URI

POST /v2.0/{project_id}/workspaces/{workspace_id}/resources/{resource_id}/service-instance

**Table 1** Path parameters
Parameter	Mandatory	Type	Description
project_id	Yes	String	Project ID. For details on how to obtain the project ID, see Obtaining a Project ID.
resource_id	Yes	String	Resource ID (data source ID or scenario ID)
workspace_id	Yes	String	Workspace ID

Request Parameters

**Table 2** Request header parameters
Parameter	Mandatory	Type	Description
Content-Type	Yes	String	Content type. The value is application/json.
X-Auth-Token	Yes	String	User token. For details on how to obtain the user token, see Obtaining a User Token Through Password Authentication.

**Table 3** Request body parameters
Parameter	Mandatory	Type	Description
job_name	Yes	String	Job name, which is a string of 1 to 64 characters and contains letters, digits, underscores (_), and hyphens (-) Minimum: 1 Maximum: 64
description	No	String	Description
category	Yes	String	Category. The option is: SERVICE indicates the online service.
job_type	Yes	String	Job type. The option is: infer indicates the inference service.
job_config	Yes	jobConfig object	Job settings
topicUrn	No	String	Notification message settings

**Table 4** jobConfig
Parameter	Mandatory	Type	Description
nearline_recall_param	No	NearLineRecallParam object	Parameters of a nearline retrieval job (required for nearline retrieval jobs)
max_recommended_num	No	Integer	Max. number of candidate sets (required for retrieval jobs) Minimum: 1 Maximum: 200
match_feature_pairs	No	Array of MatchFeaturePair objects	Matched feature pair (required for feature matching-based retrieval jobs)
striping	No	Striping object	Row-based strategy (required for feature matching-based retrieval job, itemCF retrieval job, and userCF retrieval job)
match_type	No	String	Matching type (required for feature matching-based retrieval jobs). The options are: UI: recommending items based on users UU: recommending users based on users II: recommending items based on items IU: recommending users based on items
matrix_factorization	No	MatrixFactorization object	Matrix decomposition parameter settings (required for the ALS-based MF jobs)
behavior_frequencys	No	Array of BehaviorFrequency objects	Behavior frequency information (required for historical behavior-based candidate set generation jobs)
file_path	No	String	File path (required for business rule - manual import retrieval jobs)
ucb_param	No	UcbParam object	UCB job parameter (required for UCB-based retrieval jobs)
behavior_gravity	No	BehaviorGravity object	Gravity decay factor (required for the comprehensive behavior popularity-based retrieval jobs)
category	No	Category object	Type (required for the comprehensive behavior popularity-based retrieval jobs)
behavior_logic	No	String	Behavior filter logic (required for historical behavior filter jobs). The options are: AND: If both conditions are met, the data is filtered. OR: If one condition is met, the data is filtered.
features_engineering	No	EtlBasicParameter object	Feature parameter (required for offline feature engineering jobs)
sample_param	No	SampleParam object	Sample parameter (required for offline feature engineering jobs)
deep_learning_parameters	No	DeepLearingParam object	General parameters of a ranking job (required by LR, DeepFM, and AutoGroup)
algorithm_specify_parameters	No	AlgorithmSpecifyParameters object	Specific parameter of a ranking algorithm (required for LR, DeepFM, and AutoGroup)
load_widetable	No	Boolean	Importing a wide table (required for offline data import jobs)
load_profile	No	Boolean	Importing a profile (required for offline data import jobs)
save_mode	No	String	Retaining a wide table (required for offline data import jobs). The options are: append: Yes new: No -Overwrite: Overwrite
indicators	No	Array of Indicator objects	Statistical indicator (required for effect evaluation jobs)
offline_rank_job_name	No	String	Name of an offline ranking job (required for online training jobs)
update_interval	No	Integer	Update interval (required for online training jobs)
optimizer	No	Optimizer object	Optimizer (required for online training jobs)
flows	No	Flow object	Online process flow (required for online training jobs)

**Table 5** NearLineRecallParam
Parameter	Mandatory	Type	Description
time_limit	No	Boolean	Time filter
timeFeature	No	String	Time feature
retainDays	No	Integer	Retention period (days)
recall_fileds	No	Array of RecallFiled objects	Retrieved field
itemCF_job_name	No	String	Name of an itemCF job

**Table 6** RecallFiled
Parameter	Mandatory	Type	Description
name	No	String	Field name
value	No	Integer	Number of used field values Minimum: 1 Maximum: 10 Default: 1

**Table 7** MatchFeaturePair
Parameter	Mandatory	Type	Description
user_feature_name	No	String	User feature
item_feature_name	No	String	Item feature
weight	No	Double	Weight
match_count	No	Boolean	Measurement of the number of matched tags

**Table 8** Striping
Parameter	Mandatory	Type	Description
nearest_neighborhood	Yes	Integer	Nearest neighbors
band	Yes	Integer	Similarity degree Minimum: 1 Maximum: 20
row	Yes	Integer	Similarity distance Minimum: 1 Maximum: 10

**Table 9** MatrixFactorization
Parameter	Mandatory	Type	Description
implicit_vector_rank	Yes	Integer	Embedding size Minimum: 1 Maximum: 100
regular_param	Yes	Double	Optimization lambda Minimum: 1.0E-8 Maximum: 1
max_iterator_num	Yes	Integer	Number of iterations Minimum: 1 Maximum: 50

**Table 10** BehaviorFrequency
Parameter	Mandatory	Type	Description
behavior_type	Yes	String	Behavior type. The options are: view indicates that you browsed an item. click indicates that you clicked an item. collect indicates that you added an item to favorites. uncollect indicates that you removed an item from favorites. search_click indicates that you searched and clicked an item. comment indicates that you made comments on an item. share indicates that you shared an item with others. like indicates that you gave an item a thumb-up. dislike indicates that you gave an item a thumb-down. grade indicates that you rated an item. consume indicates that you bought an item (primarily refers to goods). use indicates that you watched videos/listened to music/read books. download indicates that you downloaded something. -tip indicates that you gave somebody or something a reward. subscribe indicates that you followed somebody or something.
lower_limit	No	Integer	Min. times Minimum: 1
upper_limit	No	Integer	Max. times Minimum: 1
time_interval	Yes	Integer	Time range Minimum: 1

**Table 11** UcbParam
Parameter	Mandatory	Type	Description
alpha	Yes	Double	Tradeoff parameter Minimum: 0 Maximum: 1
min_used_num	Yes	Integer	Min. number of behaviors Minimum: 30 Maximum: 1000

**Table 12** BehaviorGravity
Parameter	Mandatory	Type	Description
weaken_factor	No	Double	Decay factor Minimum: 0.1 Maximum: 5
view_type	No	String	Behavior quantity counting mode. The options are: pv: page views or clicks, measuring the number of web pages visited by website users. uv: unique visitors, counting the number of individual users accessing a site within the reporting period.
algo_type	No	String	Algorithm type. The options are: normal time

**Table 13** Category
Parameter	Mandatory	Type	Description
user_meta_list	No	Array of strings	User feature
item_meta_list	No	Array of strings	Item feature

**Table 14** EtlBasicParameter
Parameter	Mandatory	Type	Description
user_features	No	Array of FeatureTransformation objects	User feature
item_features	No	Array of FeatureTransformation objects	Item feature
rank_etl_filter	No	RankETLFilter object	Filter parameter

**Table 15** FeatureTransformation
Parameter	Mandatory	Type	Description
attr	No	Attribute object	Feature
discrete_method	No	String	Discrete method. The options are: equal_distance_discrete indicates equal-distance discretization. user_define_discrete indicates custom discrete distance. -normalize indicates normalization. null indicates non-discretization.
params	No	Object	Specific processing parameter

**Table 16** Attribute
Parameter	Mandatory	Type	Description
name	Yes	String	Name
data_type	No	String	Data type
other_uses	No	Array of strings	Other usage

**Table 17** RankETLFilter
Parameter	Mandatory	Type	Description
filter_type	Yes	String	Behavior deduplication mode. The options are: abs_weight indicates the absolute weight value. date indicates the date.
time_type	Yes	String	Time type. The options are: Day Week Month
is_monday_first	No	Boolean	Whether Monday is the first day

**Table 18** SampleParam
Parameter	Mandatory	Type	Description
divide_type	Yes	String	Division mode of training and test sets. The options are: TIME indicates the time ratio. RAMDOM indicates the number ratio.
train_rate	No	Double	Training data ratio Minimum: 0.01 Maximum: 1
test_rate	No	Double	Test data ratio Minimum: 0.01 Maximum: 1

**Table 19** DeepLearingParam
Parameter	Mandatory	Type	Description
initial_parameters	No	Initial object	Initialization parameter
optimize_parameters	No	Optimizer object	Optimization parameter
regular_parameters	No	Regular object	Lambda parameter
max_iterations	No	Integer	Max. iterations Minimum: 1 Maximum: 1000
early_stop_iterations	No	Integer	Iterations at early stopping Minimum: 1 Maximum: 1000
batch_size	No	Integer	Batch size Minimum: 1
dataset_split_parts	No	Integer	Number of training datasets to be split Minimum: 1 Maximum: 10
restart_train	No	Boolean	Retraining

**Table 20** Initial
Parameter	Mandatory	Type	Description
initial_method	Yes	String	Initialization method Enumeration values: normal uniform xavier
mean_value	No	Double	Mean Minimum: -1 Maximum: 1
standard_deviation	No	Double	Standard deviation Minimum: 0 Maximum: 1
min_value	No	Double	Min. value Minimum: -1 Maximum: 0
max_value	No	Double	Max. value Minimum: 0 Maximum: 1

**Table 21** Regular
Parameter	Mandatory	Type	Description
l2_regularization	No	Double	Lambda 2 Minimum: 0 Maximum: 1
regular_loss_compute_mode	No	String	Regular loss calculation mode Enumeration values: full batch
embed_l2_regularization	No	Double	Lambda 2 of embedding size Minimum: 0 Maximum: 1
wide_l2_regularization	No	Double	Lambda 2 of the wide part Minimum: 0 Maximum: 1
structure_l2_regularization	No	Double	Lambda 2 of the structured part Minimum: 0 Maximum: 1

**Table 22** AlgorithmSpecifyParameters
Parameter	Mandatory	Type	Description
latent_vector_length	No	Integer	Embedding size (required for DeepFM) Minimum: 1 Maximum: 100
architecture	No	Array of integers	Neural network structure (required for DeepFM)
active_function	No	String	Activation function (required for DeepFM and AutoGroup) Enumeration values: relu sigmoid tanh
value_keep_probability	No	Double	Neuron retention probability (required for DeepFM and AutoGroup) Minimum: 0 Maximum: 1
embed_size	No	Array of integers	Embedding size of each degree (required for AutoGroup)
mlp_architecture	No	Array of integers	Neural network structure (required for AutoGroup)
max_order	No	Integer	Max. interactions (required for AutoGroup)
hash_sizes	No	Array of integers	Hash length (required for AutoGroup)
hash_compensation	No	Array of numbers	Feature interaction penalty coefficient (required for AutoGroup)
use_wide_part	No	Boolean	Wide part required (required for AutoGroup)
structure_optimizer	No	Optimizer object	Optimizer parameter (required for AutoGroup)
merge_multi_hot	No	Boolean	Merge multi-value feature (required for AutoGroup)
fix_structure	No	Boolean	Fix hash structure (required for AutoGroup)

**Table 23** Indicator
Parameter	Mandatory	Type	Description
indicator_name	No	String	Indicator. The options are: PV UV Custom
indicator_params	No	IndicatorParam object	Indicator parameter (required for custom metric)

**Table 24** IndicatorParam
Parameter	Mandatory	Type	Description
customize_parameter	No	CustomizeParameter object	Custom parameter
customize_formula	No	CustomizeFormula object	Custom formula

**Table 25** CustomizeParameter
Parameter	Mandatory	Type	Description
alias	Yes	String	Alias
behavior_type	Yes	String	Behavior type
threshold	No	Double	Threshold Minimum: 0 Maximum: 1
deduplication	Yes	String	Deduplication

**Table 26** CustomizeFormula
Parameter	Mandatory	Type	Description
alias	No	String	Alias
formula	No	String	Formula

**Table 27** Optimizer
Parameter	Mandatory	Type	Description
type	No	String	Optimizer type Enumeration values: adam adagrad ftrl
learning_rate	No	Integer	Learning rate Minimum: 0 Maximum: 1
initial_accumulator_value	No	Double	Initial gradient sum Minimum: 0 Maximum: 1
lambda1	No	Double	Lambda 1 Minimum: 0 Maximum: 1
lambda2	No	Double	Lambda 2 Minimum: 0 Maximum: 1
epsilon	No	Double	Epsilon Minimum: 0 Maximum: 1
decay_rate	No	Double	Decay factor Minimum: 0 Maximum: 1
decay_steps	No	Double	Decay step Minimum: 1

**Table 28** Flow
Parameter	Mandatory	Type	Description
flow_id	Yes	String	Process flow ID
attr_pair_rules_filter	No	Array of AttrPairRules objects	Feature pair filter
attr_pair_rules_reserve	No	Array of AttrPairRules objects	Feature pair to be reserved
deduplication_list	No	Array of Deduplication objects	Feature deduplication
attribute_info	No	AttributeInfo object	Comprehensive ranking information
bloom_filter_conf	No	BloomFilterConf object	Bloom filter settings
group_attr	No	String	The scatter attribute for grouping
pre_deal	No	Boolean	Deduplication before ranking
rank_setting	No	String	Ranking configuration information
rules	No	Rule object	Candidate set merging
filter_sets	No	Array of strings	Filter configuration information
attr_value_rules_filter	No	Array of AttrValueRules objects	Feature filter
attr_value_rules_reserve	No	Array of AttrValueRules objects	Feature to be reserved
ctr_job	No	String	Ranking job (required when click-through rate is used)
ratio	No	Integer	Traffic proportion Minimum: 1 Maximum: 100
toppings	No	Array of strings	List of candidate sets to be pinned on top

**Table 29** AttrPairRules
Parameter	Mandatory	Type	Description
attr_pairs	No	Array of AttrPair objects	Feature pair

**Table 30** AttrPair
Parameter	Mandatory	Type	Description
party_a	No	String	Feature name of the recommended item
party_b	No	String	Feature name of the recommended item

**Table 31** Deduplication
Parameter	Mandatory	Type	Description
attributes	No	Array of strings	Feature

**Table 32** AttributeInfo
Parameter	Mandatory	Type	Description
rank_feature_pairs	No	Array of RankFeaturePair objects	Matched feature pair
numerical_attrs	No	Array of NumericalAttr objects	Feature weight
num_statistics_type	No	String	Statistics mode. The options are: ORDER indicates that weights are accumulated based on the feature value sequence. ABS indicates that weights are accumulated based on the absolute value. Enumeration values: ORDER ABS

**Table 33** RankFeaturePair
Parameter	Mandatory	Type	Description
feature_name_a	No	String	Feature of the item to be recommended
feature_name_b	No	String	Feature of the recommended item
weight	No	Float	Weight Minimum: 0.01 Maximum: 1

**Table 34** NumericalAttr
Parameter	Mandatory	Type	Description
name	Yes	String	Feature name
weight	Yes	Float	Weight Minimum: 0.001 Maximum: 1

**Table 35** BloomFilterConf
Parameter	Mandatory	Type	Description
behaviors	No	Array of strings	Type of the behavior to be filtered
interval	No	Integer	Filter time Minimum: 1 Maximum: 7

**Table 36** Rule
Parameter	Mandatory	Type	Description
table_name	Yes	String	Table name of a candidate set
rule_ratio	Yes	Integer	Rule ratio Minimum: 1 Maximum: 100
priority	Yes	Integer	Priority Minimum: 1 Maximum: 10

**Table 37** AttrValueRules
Parameter	Mandatory	Type	Description
attr_values_a	No	Array of AttrValue objects	Feature value settings of the recommended item
attr_values_b	Yes	Array of AttrValue objects	Feature value settings of the item to be recommended

**Table 38** AttrValue
Parameter	Mandatory	Type	Description
name	Yes	String	Feature name
value	Yes	String	Feature value

Response Parameters

Status code: 200

**Table 39** Response body parameters
Parameter	Type	Description
is_success	Boolean	Whether the request is successful
job	jobs object	Service description
message	String	Response message (This field is not returned when the request is successful.)
error_code	String	Error code (This parameter is not returned when the request is successful.)

**Table 40** jobs
Parameter	Type	Description
category	String	Category
config_info	String	Configuration information
description	String	Description
job_id	String	Job ID
job_name	String	Job name
job_type	String	Job type
next_schedule_time	Integer	Next scheduling time
platform	String	Platform
resource_id	String	Resource ID
schedule	String	Scheduling parameter
status	String	Status
workspace_id	String	Workspace ID
job_config	jobConfig object	Job settings

**Table 41** jobConfig
Parameter	Type	Description
nearline_recall_param	NearLineRecallParam object	Parameters of a nearline retrieval job (required for nearline retrieval jobs)
max_recommended_num	Integer	Max. number of candidate sets (required for retrieval jobs) Minimum: 1 Maximum: 200
match_feature_pairs	Array of MatchFeaturePair objects	Matched feature pair (required for feature matching-based retrieval jobs)
striping	Striping object	Row-based strategy (required for feature matching-based retrieval job, itemCF retrieval job, and userCF retrieval job)
match_type	String	Matching type (required for feature matching-based retrieval jobs). The options are: UI: recommending items based on users UU: recommending users based on users II: recommending items based on items IU: recommending users based on items
matrix_factorization	MatrixFactorization object	Matrix decomposition parameter settings (required for the ALS-based MF jobs)
behavior_frequencys	Array of BehaviorFrequency objects	Behavior frequency information (required for historical behavior-based candidate set generation jobs)
file_path	String	File path (required for business rule - manual import retrieval jobs)
ucb_param	UcbParam object	UCB job parameter (required for UCB-based retrieval jobs)
behavior_gravity	BehaviorGravity object	Gravity decay factor (required for the comprehensive behavior popularity-based retrieval jobs)
category	Category object	Type (required for the comprehensive behavior popularity-based retrieval jobs)
behavior_logic	String	Behavior filter logic (required for historical behavior filter jobs). The options are: AND: If both conditions are met, the data is filtered. OR: If one condition is met, the data is filtered.
features_engineering	EtlBasicParameter object	Feature parameter (required for offline feature engineering jobs)
sample_param	SampleParam object	Sample parameter (required for offline feature engineering jobs)
deep_learning_parameters	DeepLearingParam object	General parameters of a ranking job (required by LR, DeepFM, and AutoGroup)
algorithm_specify_parameters	AlgorithmSpecifyParameters object	Specific parameter of a ranking algorithm (required for LR, DeepFM, and AutoGroup)
load_widetable	Boolean	Importing a wide table (required for offline data import jobs)
load_profile	Boolean	Importing a profile (required for offline data import jobs)
save_mode	String	Retaining a wide table (required for offline data import jobs). The options are: append: Yes new: No -Overwrite: Overwrite
indicators	Array of Indicator objects	Statistical indicator (required for effect evaluation jobs)
offline_rank_job_name	String	Name of an offline ranking job (required for online training jobs)
update_interval	Integer	Update interval (required for online training jobs)
optimizer	Optimizer object	Optimizer (required for online training jobs)
flows	Flow object	Online process flow (required for online training jobs)

**Table 42** NearLineRecallParam
Parameter	Type	Description
time_limit	Boolean	Time filter
timeFeature	String	Time feature
retainDays	Integer	Retention period (days)
recall_fileds	Array of RecallFiled objects	Retrieved field
itemCF_job_name	String	Name of an itemCF job

**Table 43** RecallFiled
Parameter	Type	Description
name	String	Field name
value	Integer	Number of used field values Minimum: 1 Maximum: 10 Default: 1

**Table 44** MatchFeaturePair
Parameter	Type	Description
user_feature_name	String	User feature
item_feature_name	String	Item feature
weight	Double	Weight
match_count	Boolean	Measurement of the number of matched tags

**Table 45** Striping
Parameter	Type	Description
nearest_neighborhood	Integer	Nearest neighbors
band	Integer	Similarity degree Minimum: 1 Maximum: 20
row	Integer	Similarity distance Minimum: 1 Maximum: 10

**Table 46** MatrixFactorization
Parameter	Type	Description
implicit_vector_rank	Integer	Embedding size Minimum: 1 Maximum: 100
regular_param	Double	Optimization lambda Minimum: 1.0E-8 Maximum: 1
max_iterator_num	Integer	Number of iterations Minimum: 1 Maximum: 50

**Table 47** BehaviorFrequency
Parameter	Type	Description
behavior_type	String	Behavior type. The options are: view indicates that you browsed an item. click indicates that you clicked an item. collect indicates that you added an item to favorites. uncollect indicates that you removed an item from favorites. search_click indicates that you searched and clicked an item. comment indicates that you made comments on an item. share indicates that you shared an item with others. like indicates that you gave an item a thumb-up. dislike indicates that you gave an item a thumb-down. grade indicates that you rated an item. consume indicates that you bought an item (primarily refers to goods). use indicates that you watched videos/listened to music/read books. download indicates that you downloaded something. -tip indicates that you gave somebody or something a reward. subscribe indicates that you followed somebody or something.
lower_limit	Integer	Min. times Minimum: 1
upper_limit	Integer	Max. times Minimum: 1
time_interval	Integer	Time range Minimum: 1

**Table 48** UcbParam
Parameter	Type	Description
alpha	Double	Tradeoff parameter Minimum: 0 Maximum: 1
min_used_num	Integer	Min. number of behaviors Minimum: 30 Maximum: 1000

**Table 49** BehaviorGravity
Parameter	Type	Description
weaken_factor	Double	Decay factor Minimum: 0.1 Maximum: 5
view_type	String	Behavior quantity counting mode. The options are: pv: page views or clicks, measuring the number of web pages visited by website users. uv: unique visitors, counting the number of individual users accessing a site within the reporting period.
algo_type	String	Algorithm type. The options are: normal time

**Table 50** Category
Parameter	Type	Description
user_meta_list	Array of strings	User feature
item_meta_list	Array of strings	Item feature

**Table 51** EtlBasicParameter
Parameter	Type	Description
user_features	Array of FeatureTransformation objects	User feature
item_features	Array of FeatureTransformation objects	Item feature
rank_etl_filter	RankETLFilter object	Filter parameter

**Table 52** FeatureTransformation
Parameter	Type	Description
attr	Attribute object	Feature
discrete_method	String	Discrete method. The options are: equal_distance_discrete indicates equal-distance discretization. user_define_discrete indicates custom discrete distance. -normalize indicates normalization. null indicates non-discretization.
params	Object	Specific processing parameter

**Table 53** Attribute
Parameter	Type	Description
name	String	Name
data_type	String	Data type
other_uses	Array of strings	Other usage

**Table 54** RankETLFilter
Parameter	Type	Description
filter_type	String	Behavior deduplication mode. The options are: abs_weight indicates the absolute weight value. date indicates the date.
time_type	String	Time type. The options are: Day Week Month
is_monday_first	Boolean	Whether Monday is the first day

**Table 55** SampleParam
Parameter	Type	Description
divide_type	String	Division mode of training and test sets. The options are: TIME indicates the time ratio. RAMDOM indicates the number ratio.
train_rate	Double	Training data ratio Minimum: 0.01 Maximum: 1
test_rate	Double	Test data ratio Minimum: 0.01 Maximum: 1

**Table 56** DeepLearingParam
Parameter	Type	Description
initial_parameters	Initial object	Initialization parameter
optimize_parameters	Optimizer object	Optimization parameter
regular_parameters	Regular object	Lambda parameter
max_iterations	Integer	Max. iterations Minimum: 1 Maximum: 1000
early_stop_iterations	Integer	Iterations at early stopping Minimum: 1 Maximum: 1000
batch_size	Integer	Batch size Minimum: 1
dataset_split_parts	Integer	Number of training datasets to be split Minimum: 1 Maximum: 10
restart_train	Boolean	Retraining

**Table 57** Initial
Parameter	Type	Description
initial_method	String	Initialization method Enumeration values: normal uniform xavier
mean_value	Double	Mean Minimum: -1 Maximum: 1
standard_deviation	Double	Standard deviation Minimum: 0 Maximum: 1
min_value	Double	Min. value Minimum: -1 Maximum: 0
max_value	Double	Max. value Minimum: 0 Maximum: 1

**Table 58** Regular
Parameter	Type	Description
l2_regularization	Double	Lambda 2 Minimum: 0 Maximum: 1
regular_loss_compute_mode	String	Regular loss calculation mode Enumeration values: full batch
embed_l2_regularization	Double	Lambda 2 of embedding size Minimum: 0 Maximum: 1
wide_l2_regularization	Double	Lambda 2 of the wide part Minimum: 0 Maximum: 1
structure_l2_regularization	Double	Lambda 2 of the structured part Minimum: 0 Maximum: 1

**Table 59** AlgorithmSpecifyParameters
Parameter	Type	Description
latent_vector_length	Integer	Embedding size (required for DeepFM) Minimum: 1 Maximum: 100
architecture	Array of integers	Neural network structure (required for DeepFM)
active_function	String	Activation function (required for DeepFM and AutoGroup) Enumeration values: relu sigmoid tanh
value_keep_probability	Double	Neuron retention probability (required for DeepFM and AutoGroup) Minimum: 0 Maximum: 1
embed_size	Array of integers	Embedding size of each degree (required for AutoGroup)
mlp_architecture	Array of integers	Neural network structure (required for AutoGroup)
max_order	Integer	Max. interactions (required for AutoGroup)
hash_sizes	Array of integers	Hash length (required for AutoGroup)
hash_compensation	Array of numbers	Feature interaction penalty coefficient (required for AutoGroup)
use_wide_part	Boolean	Wide part required (required for AutoGroup)
structure_optimizer	Optimizer object	Optimizer parameter (required for AutoGroup)
merge_multi_hot	Boolean	Merge multi-value feature (required for AutoGroup)
fix_structure	Boolean	Fix hash structure (required for AutoGroup)

**Table 60** Indicator
Parameter	Type	Description
indicator_name	String	Indicator. The options are: PV UV Custom
indicator_params	IndicatorParam object	Indicator parameter (required for custom metric)

**Table 61** IndicatorParam
Parameter	Type	Description
customize_parameter	CustomizeParameter object	Custom parameter
customize_formula	CustomizeFormula object	Custom formula

**Table 62** CustomizeParameter
Parameter	Type	Description
alias	String	Alias
behavior_type	String	Behavior type
threshold	Double	Threshold Minimum: 0 Maximum: 1
deduplication	String	Deduplication

**Table 63** CustomizeFormula
Parameter	Type	Description
alias	String	Alias
formula	String	Formula

**Table 64** Optimizer
Parameter	Type	Description
type	String	Optimizer type Enumeration values: adam adagrad ftrl
learning_rate	Integer	Learning rate Minimum: 0 Maximum: 1
initial_accumulator_value	Double	Initial gradient sum Minimum: 0 Maximum: 1
lambda1	Double	Lambda 1 Minimum: 0 Maximum: 1
lambda2	Double	Lambda 2 Minimum: 0 Maximum: 1
epsilon	Double	Epsilon Minimum: 0 Maximum: 1
decay_rate	Double	Decay factor Minimum: 0 Maximum: 1
decay_steps	Double	Decay step Minimum: 1

**Table 65** Flow
Parameter	Type	Description
flow_id	String	Process flow ID
attr_pair_rules_filter	Array of AttrPairRules objects	Feature pair filter
attr_pair_rules_reserve	Array of AttrPairRules objects	Feature pair to be reserved
deduplication_list	Array of Deduplication objects	Feature deduplication
attribute_info	AttributeInfo object	Comprehensive ranking information
bloom_filter_conf	BloomFilterConf object	Bloom filter settings
group_attr	String	The scatter attribute for grouping
pre_deal	Boolean	Deduplication before ranking
rank_setting	String	Ranking configuration information
rules	Rule object	Candidate set merging
filter_sets	Array of strings	Filter configuration information
attr_value_rules_filter	Array of AttrValueRules objects	Feature filter
attr_value_rules_reserve	Array of AttrValueRules objects	Feature to be reserved
ctr_job	String	Ranking job (required when click-through rate is used)
ratio	Integer	Traffic proportion Minimum: 1 Maximum: 100
toppings	Array of strings	List of candidate sets to be pinned on top

**Table 66** AttrPairRules
Parameter	Type	Description
attr_pairs	Array of AttrPair objects	Feature pair

**Table 67** AttrPair
Parameter	Type	Description
party_a	String	Feature name of the recommended item
party_b	String	Feature name of the recommended item

**Table 68** Deduplication
Parameter	Type	Description
attributes	Array of strings	Feature

**Table 69** AttributeInfo
Parameter	Type	Description
rank_feature_pairs	Array of RankFeaturePair objects	Matched feature pair
numerical_attrs	Array of NumericalAttr objects	Feature weight
num_statistics_type	String	Statistics mode. The options are: ORDER indicates that weights are accumulated based on the feature value sequence. ABS indicates that weights are accumulated based on the absolute value. Enumeration values: ORDER ABS

**Table 70** RankFeaturePair
Parameter	Type	Description
feature_name_a	String	Feature of the item to be recommended
feature_name_b	String	Feature of the recommended item
weight	Float	Weight Minimum: 0.01 Maximum: 1

**Table 71** NumericalAttr
Parameter	Type	Description
name	String	Feature name
weight	Float	Weight Minimum: 0.001 Maximum: 1

**Table 72** BloomFilterConf
Parameter	Type	Description
behaviors	Array of strings	Type of the behavior to be filtered
interval	Integer	Filter time Minimum: 1 Maximum: 7

**Table 73** Rule
Parameter	Type	Description
table_name	String	Table name of a candidate set
rule_ratio	Integer	Rule ratio Minimum: 1 Maximum: 100
priority	Integer	Priority Minimum: 1 Maximum: 10

**Table 74** AttrValueRules
Parameter	Type	Description
attr_values_a	Array of AttrValue objects	Feature value settings of the recommended item
attr_values_b	Array of AttrValue objects	Feature value settings of the item to be recommended

**Table 75** AttrValue
Parameter	Type	Description
name	String	Feature name
value	String	Feature value

Example Requests

This API is used to publish an inference service.

/v2.0/testuuidxxxxxxxxxxxxxxxxxxxxxxxx/workspaces/testuuidxxxxxxxxxxxxxxxxxxxxxxxx/resources/testuuidxxxxxxxxxxxxxxxxxxxxxxxx/service-instance

{
  "category" : "SERVICE",
  "job_type" : "infer",
  "job_config" : {
    "flows" : [ {
      "ratio" : 100,
      "rules" : [ {
        "table_name" : "item-UIREC",
        "rule_ratio" : 100,
        "priority" : "1"
      } ],
      "flow_id" : "flow1"
    } ]
  },
  "description" : "inference service",
  "job_name" : "online1"
}

Example Responses

Status code: 200

{
  "job" : {
    "category" : "SERVICE",
    "job_name" : "online1",
    "job_id" : "testuuidxxxxxxxxxxxxxxxxxxxxxxxx",
    "description" : "inference service",
    "job_type" : "infer",
    "status" : "Draft",
    "platform" : "AIP",
    "workspace_id" : "testuuidxxxxxxxxxxxxxxxxxxxxxxxx",
    "resource_id" : "testuuidxxxxxxxxxxxxxxxxxxxxxxxx"
  },
  "is_success" : true
}

Status Codes

Status Code	Description
200	OK

Error Codes

See Error Codes.

Parent topic: Online Service

Last Article: Online Service

Next Article: Querying Details About an Online Service

Did this article solve your problem?

Thank you for your score！Your feedback would help us improve the website.

Products

Compute

Application

Dedicated Cloud

Storage

Management & Deployment

Migration

Network

Enterprise Intelligence

Video

Database

Edge Cloud Services

DevCloud

Security

Cloud Communications

Internet of Things

Solutions

Industry-Specific Solutions

General-Purpose Solutions

Security

DevOps

Enterprise Intelligence

Essential Platform

Big Data

Visual Cognition

Speech and Semantics

Support

Help Center

Customer Services

Developers

Console

语言 - Language

中国站 - 简体中文

中国站 - English

International - 简体中文

International - English