Network Analysis

Treats offers as nodes in a directed graph. Edges encode how often one offer is accepted when another is co-presented. Personalized PageRank (or a related centrality pass) produces scores that highlight hub offers and structures implied by co-occurrence and acceptance.

Algorithm

Config value: "approach": "Network" (no behaviorAlgos sub-approach)

Edge weight from offer \(i\) to offer \(j\):

\(w(i \rightarrow j) = \frac{\text{accepted\_count}_j}{\text{total\_count}_j}\)

(Intuition: mass flows toward \(j\) in proportion to how often \(j\) is accepted relative to presentations involving the co-occurrence context.)

PageRank-style update (schematic):

\(\mathrm{PR}(u) = (1 - \alpha) \cdot \mathrm{personalization}(u) + \alpha \cdot \sum_{v} \mathrm{PR}(v) \cdot \frac{w(v \rightarrow u)}{\sum_{k} w(v \rightarrow k)}\)

• \(\alpha\): damping / teleport mixing (implementation default may use \(\alpha = 1.0\) for a specific variant—verify against your runtime version).
• iterations: Often set to \(N\) = number of offers for convergence passes over the graph.

Parameters

• alpha (\(\alpha\)): Damping factor between random jumps and graph propagation. Default: 1.0 (check runtime notes for your build).
• iterations: Number of propagation iterations; default ties to offer count (\(N\)).
• Processing Window: Time window in milliseconds for historical data used to build the graph.
• Historical Count: Max records to process per update cycle.

Cold Start

Recommendations are always returned. The real-time training path always produces a scored options array:

• No history: Every offer in the options store receives a uniform random score. All offers are ranked and passed to the post-score class.
• Sparse co-occurrence data: The algorithm may not produce meaningful graph-based rankings. Unscored offers receive a random fallback score.
• Sufficient co-occurrence data: PageRank scores drive the ranking, highlighting hub offers and structural relationships in the acceptance network.

The scored options are then sorted by arm_reward and handed to the configured dynamic post-score class, which controls the final offer selection and response formatting.

When To Use

• Offer relationships matter: bundles, complements, sequences, or co-presented sets
• Identifying hub or structurally central offers in an acceptance network

When NOT To Use

• Offers are independent (no meaningful co-presentation signal)
• You need fast exploration of brand-new items without a graph (prefer bandit methods)

Example

from prediction.apis import deployment_management as dm
from prediction.apis import online_learning_management as ol
from prediction import jwt_access
 
auth = jwt_access.Authenticate("http://localhost:3001/api", ecosystem_username, ecosystem_password)
 
deployment_id = "demo-network-analysis"
 
online_learning_uuid = ol.create_online_learning(
        auth,
        algorithm="ecosystem_rewards",
        name=deployment_id,
        description="Network Analysis (PageRank) configuration",
        feature_store_collection="set_up_features",
        feature_store_database="my_mongo_database",
        options_store_database="my_mongo_database",
        options_store_collection="demo-deployment_options",
        randomisation_processing_count=5000,
        randomisation_processing_window=604800000,
        contextual_variables_offer_key="offer",
        create_options_index=True,
        create_covering_index=True
)
 
online_learning = dm.define_deployment_multi_armed_bandit(epsilon=0, dynamic_interaction_uuid=online_learning_uuid)
 
parameter_access = dm.define_deployment_parameter_access(
    auth,
    lookup_key="customer_id",
    lookup_type="string",
    database="my_mongo_database",
    table_collection="customer_feature_store",
    datasource="mongodb"
)
 
deployment_step = dm.create_deployment(
    auth,
    project_id="demo-project",
    deployment_id=deployment_id,
    description="Network Analysis demo deployment",
    version="001",
    plugin_post_score_class="PlatformDynamicEngagement.java",
    plugin_pre_score_class="PreScoreDynamic.java",
    scoring_engine_path_dev="http://localhost:8091",
    mongo_connect=f"mongodb://{mongo_user}:{mongo_password}@localhost:54445/?authSource=admin",
    parameter_access=parameter_access,
    multi_armed_bandit=online_learning
)