Use the fitpredict function in Interpolate and InterpolateNeighbors

Project.toml

@@ -30,7 +30,7 @@ ColumnSelectors = "0.1"
 Combinatorics = "1.0"
 DataScienceTraits = "0.1"
 Distances = "0.10"
-GeoStatsModels = "0.1"
+GeoStatsModels = "0.2"
 GeoTables = "1.9"
 Meshes = "0.35"
 TableDistances = "0.3"

src/GeoStatsTransforms.jl

@@ -23,7 +23,7 @@ using LinearAlgebra
 using Statistics
 
 using Unitful: AffineQuantity
-using GeoStatsModels: GeoStatsModel, fit, predict, predictprob
+using GeoStatsModels: GeoStatsModel, fitpredict
 using ColumnSelectors: ColumnSelector, SingleColumnSelector
 using ColumnSelectors: Column, AllSelector, NoneSelector
 using ColumnSelectors: selector, selectsingle

src/interpneighbors.jl

@@ -88,12 +88,11 @@ InterpolateNeighbors(domain, pairs::Pair{<:Any,<:GeoStatsModel}...; kwargs...) =
 isrevertible(::Type{<:InterpolateNeighbors}) = false
 
 function apply(transform::InterpolateNeighbors, geotable::AbstractGeoTable)
-  dom = domain(geotable)
   tab = values(geotable)
   cols = Tables.columns(tab)
   vars = Tables.columnnames(cols)
 
-  idom = transform.domain
+  domain = transform.domain
   selectors = transform.selectors
   models = transform.models
   minneighbors = transform.minneighbors
@@ -103,73 +102,13 @@ function apply(transform::InterpolateNeighbors, geotable::AbstractGeoTable)
   point = transform.point
   prob = transform.prob
 
-  nobs = nelements(dom)
-  if maxneighbors > nobs || maxneighbors < 1
-    @warn "Invalid maximum number of neighbors. Adjusting to $nobs..."
-    maxneighbors = nobs
-  end
-
-  if minneighbors > maxneighbors || minneighbors < 1
-    @warn "Invalid minimum number of neighbors. Adjusting to 1..."
-    minneighbors = 1
-  end
-
-  data = if point
-    pset = PointSet(centroid(dom, i) for i in 1:nobs)
-    _adjustunits(georef(values(geotable), pset))
-  else
-    _adjustunits(geotable)
-  end
-
-  # preprocess variable models
-  varmodels = mapreduce(vcat, selectors, models) do selector, model
+  interps = map(selectors, models) do selector, model
     svars = selector(vars)
-    [var => model for var in svars]
+    data = geotable[:, svars]
+    fitpredict(model, data, domain; point, prob, minneighbors, maxneighbors, neighborhood, distance)
   end
 
-  # determine bounded search method
-  searcher = searcher_ui(dom, maxneighbors, distance, neighborhood)
-
-  # pre-allocate memory for neighbors
-  neighbors = Vector{Int}(undef, maxneighbors)
-
-  # prediction order
-  inds = traverse(idom, LinearPath())
-
-  # predict variable values
-  function pred(var, model)
-    map(inds) do ind
-      # centroid of estimation
-      center = centroid(idom, ind)
-
-      # find neighbors with data
-      nneigh = search!(neighbors, center, searcher)
-
-      # predict if enough neighbors
-      if nneigh ≥ minneighbors
-        # final set of neighbors
-        ninds = view(neighbors, 1:nneigh)
-
-        # view neighborhood with data
-        samples = view(data, ninds)
-
-        # fit model to data
-        fmodel = fit(model, samples)
-
-        # save prediction
-        geom = point ? center : idom[ind]
-        pfun = prob ? predictprob : predict
-        pfun(fmodel, var, geom)
-      else # missing prediction
-        missing
-      end
-    end
-  end
-
-  pairs = (var => pred(var, model) for (var, model) in varmodels)
-  newtab = (; pairs...) |> Tables.materializer(tab)
-
-  newgeotable = georef(newtab, idom)
+  newgeotable = reduce(hcat, interps)
 
   newgeotable, nothing
 end

src/interpolate.jl

@@ -45,47 +45,23 @@ Interpolate(domain, pairs::Pair{<:Any,<:GeoStatsModel}...; kwargs...) =
 isrevertible(::Type{<:Interpolate}) = false
 
 function apply(transform::Interpolate, geotable::AbstractGeoTable)
-  dom = domain(geotable)
   tab = values(geotable)
   cols = Tables.columns(tab)
   vars = Tables.columnnames(cols)
 
-  idom = transform.domain
+  domain = transform.domain
   selectors = transform.selectors
   models = transform.models
   point = transform.point
   prob = transform.prob
 
-  data = if point
-    pset = PointSet(centroid(dom, i) for i in 1:nelements(dom))
-    _adjustunits(georef(values(geotable), pset))
-  else
-    _adjustunits(geotable)
-  end
-
-  # preprocess variable models
-  varmodels = mapreduce(vcat, selectors, models) do selector, model
-    fmodel = fit(model, data)
+  interps = map(selectors, models) do selector, model
     svars = selector(vars)
-    [var => fmodel for var in svars]
-  end
-
-  # prediction order
-  inds = traverse(idom, LinearPath())
-
-  # predict variable values
-  function pred(var, fmodel)
-    map(inds) do ind
-      geom = point ? centroid(idom, ind) : idom[ind]
-      pfun = prob ? predictprob : predict
-      pfun(fmodel, var, geom)
-    end
+    data = geotable[:, svars]
+    fitpredict(model, data, domain; point, prob, neighbors=false)
   end
 
-  pairs = (var => pred(var, fmodel) for (var, fmodel) in varmodels)
-  newtab = (; pairs...) |> Tables.materializer(tab)
-
-  newgeotable = georef(newtab, idom)
+  newgeotable = reduce(hcat, interps)
 
   newgeotable, nothing
 end

src/utils.jl

@@ -2,22 +2,6 @@
 # Licensed under the MIT License. See LICENSE in the project root.
 # ------------------------------------------------------------------
 
-"""
-    searcher_ui(domain, maxneighbors, distance, neighborhood)
-
-Return the appropriate search method over the `domain` based on
-end-user inputs such as `maxneighbors`, `distance` and `neighborhood`.
-"""
-function searcher_ui(domain, maxneighbors, distance, neighborhood)
-  if isnothing(neighborhood)
-    # nearest neighbor search with a metric
-    KNearestSearch(domain, maxneighbors; metric=distance)
-  else
-    # neighbor search with ball neighborhood
-    KBallSearch(domain, maxneighbors, neighborhood)
-  end
-end
-
 #-------------
 # AGGREGATION
 #-------------