Given a set of objects (e.g. rows of a table) with scores given by different scoring methods (e.g. columns), how to rank the objects? This problem is known in the literature as the rank aggregation problem.

The problem is trivial when there is only one score for each object (one column), but ranking objects on the basis of multiple (conflicting) scores is challenging. This package provides algorithms to aggregate multiple scores stored in a tabular format (see Tables.jl) into a final rank vector.

Get the latest stable release with Julia's package manager:

] add RankAggregation

Given a table with scores score1 and score2 for objects a, b, and c:

julia> using DataFrames
julia> using RankAggregation

julia> objects = DataFrame(object=[:a,:b,:c], score1=[0.9, 0.7, 0.5], score2=[0.8, 0.9, 0.4])
3×3 DataFrame
│ Row │ object │ score1  │ score2  │
│     │ Symbol │ Float64 │ Float64 │
├─────┼────────┼─────────┼─────────┤
│ 1   │ a      │ 0.9     │ 0.8     │
│ 2   │ b      │ 0.7     │ 0.9     │
│ 3   │ c      │ 0.5     │ 0.4     │

rank the objects using:

julia> rank(objects, :score1)
3-element Array{Int64,1}:
 1
 2
 3

 julia> rank(objects, :score2)
3-element Array{Int64,1}:
 2
 1
 3

 julia> rank(objects, (:score1,:score2))
3-element Array{Int64,1}:
 1
 2
 3

Optionally, specify the aggregation method:

julia> rank(objects, (:score1,:score2), TauModel())
3-element Array{Int64,1}:
 1
 2
 3

and the reverse option:

julia> rank(objects, (:score1,:score2), rev=true)
3-element Array{Int64,1}:
 3
 2
 1

Contributions are very welcome, as are feature requests and suggestions.

Please open an issue if you encounter any problems.