Unify predict and ranking for simplicity

since `ranking` is used as a simple wrapper of `predict`.

src/base_recommender.jl

@@ -1,5 +1,5 @@
 export Recommender
-export isdefined, validate, fit!, recommend, predict, ranking
+export isdefined, validate, fit!, recommend, predict
 
 abstract type Recommender end
 
@@ -31,7 +31,7 @@ end
 function recommend(recommender::Recommender, u::Integer, k::Integer, candidates::Array{T}) where {T<:Integer}
     d = Dict{T,AbstractFloat}()
     for candidate in candidates
-        score = ranking(recommender, u, candidate)
+        score = predict(recommender, u, candidate)
         if isnan(score); continue; end
         d[candidate] = score
     end
@@ -42,9 +42,3 @@ end
 function predict(recommender::Recommender, u::Integer, i::Integer)
     error("predict is not implemented for recommender type $(typeof(recommender))")
 end
-
-# Return a ranking score of item i for user u
-function ranking(recommender::Recommender, u::Integer, i::Integer)
-    validate(recommender)
-    predict(recommender, u, i)
-end

src/baseline/co_occurrence.jl

@@ -35,7 +35,7 @@ function fit!(recommender::CoOccurrence)
     recommender.scores[:] = CC / c * 100.0
 end
 
-function ranking(recommender::CoOccurrence, u::Integer, i::Integer)
+function predict(recommender::CoOccurrence, u::Integer, i::Integer)
     validate(recommender)
     recommender.scores[i]
 end

src/baseline/most_popular.jl

@@ -22,7 +22,7 @@ function fit!(recommender::MostPopular)
     recommender.scores[:] = vec(sum(!iszero, recommender.data.R, dims=1))
 end
 
-function ranking(recommender::MostPopular, u::Integer, i::Integer)
+function predict(recommender::MostPopular, u::Integer, i::Integer)
     validate(recommender)
     recommender.scores[i]
 end

src/baseline/threshold_percentage.jl

@@ -28,7 +28,7 @@ function fit!(recommender::ThresholdPercentage)
     recommender.scores[:] = vec(users_rated_higher ./ users_rated * 100.0)
 end
 
-function ranking(recommender::ThresholdPercentage, u::Integer, i::Integer)
+function predict(recommender::ThresholdPercentage, u::Integer, i::Integer)
     validate(recommender)
     recommender.scores[i]
 end

test/baseline/test_co_occurrence.jl

@@ -1,9 +1,9 @@
 function test_co_occurrence(data)
     recommender = CoOccurrence(data, 1)
     fit!(recommender)
-    @test ranking(recommender, 1, 1) == 100.0
-    @test ranking(recommender, 1, 2) == 50.0
-    @test ranking(recommender, 1, 3) == 0.0
+    @test predict(recommender, 1, 1) == 100.0
+    @test predict(recommender, 1, 2) == 50.0
+    @test predict(recommender, 1, 3) == 0.0
 end
 
 println("-- Testing CoOccurrence recommender")

test/baseline/test_most_popular.jl

@@ -4,23 +4,23 @@ function test_most_popular()
     data = DataAccessor([1 2 3; 4 5 nothing])
     recommender = MostPopular(data)
     fit!(recommender)
-    @test ranking(recommender, 1, 1) == 2.0
-    @test ranking(recommender, 1, 3) == 1.0
+    @test predict(recommender, 1, 1) == 2.0
+    @test predict(recommender, 1, 3) == 1.0
 
     data = DataAccessor(sparse([1 2 3; 4 5 0]))
     recommender = MostPopular(data)
     fit!(recommender)
-    @test ranking(recommender, 1, 1) == 2.0
-    @test ranking(recommender, 1, 3) == 1.0
+    @test predict(recommender, 1, 1) == 2.0
+    @test predict(recommender, 1, 3) == 1.0
 
     n_users, n_items = 5, 10
     events = [Event(1, 2, 1), Event(3, 2, 1), Event(2, 6, 4)]
     data = DataAccessor(events, n_users, n_items)
     recommender = MostPopular(data)
     fit!(recommender)
-    @test ranking(recommender, 1, 1) == 0.0
-    @test ranking(recommender, 1, 2) == 2.0
-    @test ranking(recommender, 1, 6) == 1.0
+    @test predict(recommender, 1, 1) == 0.0
+    @test predict(recommender, 1, 2) == 2.0
+    @test predict(recommender, 1, 6) == 1.0
 end
 
 test_most_popular()

test/baseline/test_threshold_percentage.jl

@@ -1,8 +1,8 @@
 function test_threshold_percentage(data)
     recommender = ThresholdPercentage(data, 2.0)
     fit!(recommender)
-    @test ranking(recommender, 1, 1) == 50.0
-    @test ranking(recommender, 1, 2) == 100.0
+    @test predict(recommender, 1, 1) == 50.0
+    @test predict(recommender, 1, 2) == 100.0
 end
 
 println("-- Testing ThresholdPercentage recommender")

test/test_base_recommender.jl

@@ -8,7 +8,7 @@ function test_not_implemented_error()
     recommender = Foo(data)
     @test_throws ErrorException fit!(recommender)
     @test_throws ErrorException predict(recommender, 1, 1)
-    @test_throws ErrorException ranking(recommender, 1, 1)
+    @test_throws ErrorException predict(recommender, 1, 1)
 end
 
 function test_not_build_error()