Query-to-Query Similarity Job

Configuration Properties

Train a collaborative filtering matrix decomposition recommender using SparkML’s Alternating Least Squares (ALS) to batch-compute query-query similarities. This can be used for items-for-query recommendations as well as queries-for-query recommendations.

Note	This job is deprecated as of Fusion 5.2.0. We recommend the use of the Query-to-Query Session-Based Similarity jobs instead for better performance and query coverage.

Configuration Properties

Property Description, Type

alwaysTrain

Force model re-training

Even if a model with this modelId exists, re-train if set true

type: boolean

default value: 'true'

gridSearchWidth

Grid Search Width

Parameter grid search to be done centered around initial parameter guesses, exponential step size, this number of steps (if <= 0, no grid search)

type: integer

default value: '1'

Spark Job ID

required

The ID for this Spark job. Used in the API to reference this job. Allowed characters: a-z, A-Z, dash (-) and underscore (_)

type: string

maxLength: 128

pattern: ^[A-Za-z0-9_\-]+$

implicitRatings

Implicit Preferences

Treat training preferences as implicit signals of interest (i.e. clicks or other actions) as opposed to explicit query ratings

type: boolean

default value: 'true'

initialAlpha

Implicit Preference Confidence

Confidence weight (between 0 and 1) to give the implicit preferences (or starting guess, if doing parameter grid search)

type: number

default value: '0.5'

initialBlocks

Training Block Size

Number of sub-matrix blocks to break the training data into (default: -1, for auto-sizing)

type: integer

default value: '-1'

initialLambda

Smoothing

Smoothing parameter to avoid overfitting (or starting guess, if doing parameter grid search). Slightly larger value needed for small data sets

type: number

default value: '0.01'

initialRank

Recommender Rank

Number of user/item factors in the recommender decomposition (or starting guess for it, if doing parameter grid search)

type: integer

default value: '100'

itemIdField

Training Collection Item Id Field

Solr field name containing stored item ids

type: string

default value: 'item_id_s'

maxTrainingIterations

Maximum Training Iterations

Maximum number of iterations to use when learning the matrix decomposition

type: integer

default value: '10'

modelCollection

Model Collection

Collection to load and store the computed model (if absent, it won't be loaded or saved)

type: string

modelId

Recommender Model ID

Identifier for the recommender model. Will be used as the unique key when storing the model in Solr.

type: string

numItemsPerQuery

Number of Items per Query to Recommend

Batch compute and store this many item recommendations per query

type: integer

default value: '10'

numSims

Number of Query Similarities to Compute

Batch compute and store this many query similarities per query

type: integer

default value: '10'

outputItemsForQueriesCollection

Items-for-query Boosting Collection

Collection to store batch-computed items-for-queries recommendations (if absent, none computed)

type: string

outputQuerySimCollection

Query-to-query Similarity Collection

required

Collection to store batch-computed query/query similarities (if absent, none computed)

type: string

popularQueryMin

Training Data Filter By Popular Items

Items must have at least this # of unique users interacting with it to go into the sample

type: integer

default value: '2'

queryField

Training Collection Query Field

Solr field name containing stored queries

type: string

default value: 'query'

randomSeed

Random Seed

Pseudorandom determinism fixed by keeping this seed constant

type: integer

default value: '13'

saveModel

Save Model in Solr

Whether we should save the computed ALS model in Solr

type: boolean

default value: 'false'

trainingCollection

Recommender Training Collection

required

Item/Query preference collection (often a signals collection or signals aggregation collection)

type: string

trainingDataFilterQuery

Training Data Filter Query

Solr query to filter training data (e.g. downsampling or selecting based on min. pref values)

type: string

default value: '*:*'

trainingDataFrameConfigOptions

Dataframe Config Options

Additional spark dataframe loading configuration options

type: object

object attributes: {
}

trainingSampleFraction

Training Data Sampling Fraction

Downsample preferences for items (bounded to at least 2) by this fraction

type: number

default value: '1.0'

exclusiveMaximum: false

maximum: 1.0

type

Spark Job Type

required

type: string

default value: 'query_similarity'

enum: { query_similarity }

weightField

Training Collection Weight Field

Solr field name containing stored weights (i.e. time decayed / position weighted counts) the item has for that query

type: string

default value: 'weight_d'