/HIN-Similarity-Survey

Primary LanguageTeX

HIN-Similarity-Survey

By Zichen Zhu & Chenhao Ma

Data Structure

Basic Classes:

  • Entity:
    • Variables:
      • entityType -> entity type
      • entityId -> entity id in this type
  • Relation:
    • Variables:
      • startEntity -> the start entity
      • endEntity -> the end entity
      • weight(default:None) -> the weight of this relation object
  • Entity Info:
    • Variables:
      • entity -> its correponding entity in HIN
      • inRelations:
        • Type: dict
        • keys: the start entity type
        • Values:
          • relIndexDict:
            • Type: dict
            • keys: the start entity id
            • Values: the indexes of these correponding relations in HIN['Relations']
          • relsNum:
            • Type: int
            • the number of relations under this relation type
      • outRelations:
        • Type: dict
        • keys: the end entity type
        • Values: (similar to the inRelations values)
    • Methods:
      • addRelation(relation, relationIndex, inRelationFlag):
        • relation: a Relation object
        • relationIndex: its correponding index in HIN['Relations']
        • inRelationFlag:
          • True -> a incoming relation
          • False -> a outgoing relation
        • function: update this entity's inRelations or outRelations according to the relation and its index

Other Variables:

  • HIN:
    • Type: dict
    • Keys: ['Entities', 'EntityTypes', 'Relations', 'RelationTypes']
  • HIN['Entities']:
    • Type: list
    • Values:
      • Type: Entity Info Class
  • HIN['EntityTypes']:
    • Type: dict
    • Keys: all possible entity types
    • Values:
      • Type: dict
      • Keys: entity id of this type
      • Values: the index of an EntityInfo object in HIN['Entities']
  • HIN['Relations']:
    • Type: list
    • Values:
      • Type: Relation Class
  • HIN['RelationTypes']:
    • Type: dict
    • Keys: "(entity type1)-(entity type2)"
    • Values:
      • Type: list
      • Values:
        • Type: the index of a relation in HIN['Relations']

To Run the Code

  • Dataset source is refered in the report.
  • python3 split.py is used to produce the train and test data based on the dataset.
  • python3 HIN.py is used to load the graph and store the graph into pickle file for further use.
  • python3 test.py is used to re-produde the results shown in our report.