ALS

Alternating Least Squares (ALS)

Uses the ALS algorithm outlined in this article for the explicit case. For the implicit case, the algorithm is outlined here

Relies on Skelter John's matrix.go package for some matrix functionality.

Usage

1. Install go get github.com/timkaye11/goRecommend/ALS

2. Download the dependencies

   • If GPM is not installed, then do: brew install gpm
   • Get the dependencies: gpm install

3. Run Code

import "github.com/timkaye11/goRecommend/ALS"

func main() {
	// For this instance, cols indicate product ID ; rows indicate user ID. Remember indexing starts at 0. 
	Q := MakeRatingMatrix([]float64{
		5, 5, 5, 0, 1,
		0, 0, 0, 4, 1,
		1, 2, 3, 3, 1,
		2, 0, 4, 1, 0,
		5, 2, 0, 1, 0}, 5, 5)

	// OR load in through a text file
	// Q := Load("path/to/file", "separator") // where separator can be a comma, tally, tab, etc...

	// Train a model with 5 factors, 10 iterations, and a lambda value of 0.01.
	// 10 iterations is usually enough to reach convergence, and a lambda val of 0.01 is acceptable.
	n_factors := 5
	n_iterations := 10
	lambda := 0.01

	// Train Model Using Explicit ALS. This means that users rated each product on a scale
	// where 0 indicates not rated
	// Prints out the final error value.
	Qhat := Train(Q, n_factors, n_iterations, lambda)
	fmt.Println(Qhat)

	// Get Prediction for a user/product pair.
	fmt.Println(Predict(Qhat, 2, 1))

	// Get top - N recommended products based off of Qhat for a given user ID.
	// Args: Original user/product matrix, trained model, N, product names.
	// Returns []string of length N & error - if applicable
	// If Product Names is nil, then returns top indices for each user. Returns in descending order.

	products := []string{"Macy Gray", "The Black Keys", "Spoon", "A Tribe Called Quest", "Kanye West"}
	fmt.Println(GetTopNRecommendations(Q, Qhat, userID, n, products))

	// Implicit. Can do 'GetTopNRecommendations' in implicit case too. 
	R := TrainImplicit(Q, 5, 10, 0.01)
	fmt.Println(Predict(R, 1, 1))

}