vectormodel

Documentation

Overview

Package vectormodel provides primitives for serving Matrix Factorization based Recommender System models. We assume that the document vectors were calculated according to the paper "Collaborative Filtering for Implicit Feedback Datasets". Given the document vectors and the list of documents consumed by a user, this package can calculate the user vectors in realtime to generate recommendations.

Index

• type DocumentScore
• type VectorModel
  • func NewVectorModel(documents map[int][]float64, confidence, regularization float64) (*VectorModel, error)
  • func (vm *VectorModel) Rank(candidates []int, seenDocs map[int]bool) (scores []float64, err error)
  • func (vm *VectorModel) Recommend(seenDocs map[int]bool, n int) (recommendations []DocumentScore, err error)

Types

type DocumentScore

type DocumentScore struct {
	DocumentID int
	Score      float64
}

DocumentScore is the result of a recommendation

type VectorModel

type VectorModel struct {
	// contains filtered or unexported fields
}

VectorModel is a struct to handle document vector space models.

func NewVectorModel

func NewVectorModel(documents map[int][]float64, confidence, regularization float64) (*VectorModel, error)

NewVectorModel creates a new VectorModel

func (*VectorModel) Rank

func (vm *VectorModel) Rank(candidates []int, seenDocs map[int]bool) (scores []float64, err error)

Rank sorts a list of candidate assets for a given user history

func (*VectorModel) Recommend

func (vm *VectorModel) Recommend(seenDocs map[int]bool, n int) (recommendations []DocumentScore, err error)

Recommend returns a list of recommendedDocs and a list of scores