✨ Introduce BalancedBagging

.github/workflows/CI.yml

@@ -20,6 +20,8 @@ jobs:
       matrix:
         version:
           - '1.8'
+          - '1'
+
         os: [ubuntu-latest, windows-latest, macOS-latest]
         arch:
           - x64

.gitignore

@@ -1,2 +1,3 @@
 /Manifest.toml
+.CondaPkg/*
 .CondaPkg

Project.toml

@@ -4,10 +4,13 @@ authors = ["Essam <[email protected]> and contributors"]
 version = "1.0.0-DEV"
 
 [deps]
+JuliaFormatter = "98e50ef6-434e-11e9-1051-2b60c6c9e899"
 MLJBase = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
 MLJModelInterface = "e80e1ace-859a-464e-9ed9-23947d8ae3ea"
+MLUtils = "f1d291b0-491e-4a28-83b9-f70985020b54"
 OrderedCollections = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 
 [compat]
 MLJBase = "0.21"
@@ -17,9 +20,10 @@ julia = "1.6"
 [extras]
 DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
 Imbalance = "c709b415-507b-45b7-9a3d-1767c89fde68"
-MLJLinearModels = "6ee0df7b-362f-4a72-a706-9e79364fb692"
 MLJModels = "d491faf4-2d78-11e9-2867-c94bc002c0b7"
+MLJLinearModels = "6ee0df7b-362f-4a72-a706-9e79364fb692"
+Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]
-test = ["Test", "Imbalance", "DataFrames", "MLJLinearModels", "MLJModels"]
+test = ["Test", "Imbalance", "DataFrames", "MLJLinearModels", "MLJModels", "Tables"]

examples/BalancedBagging.ipynb

@@ -0,0 +1,165 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "ENV[\"JULIA_PKG_SERVER\"] = \"\"\n",
+    "using Pkg\n",
+    "Pkg.activate(@__DIR__)\n",
+    "Pkg.instantiate()\n",
+    "\n",
+    "\n",
+    "using MLJBalancing\n",
+    "using Imbalance\n",
+    "using MLJ\n",
+    "using Random"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Load Data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "((Column1 = [0.9695150609084499, 0.012898301755861596, 0.7555027304121053, 0.3467415729179013, 0.35969402837473463, 0.2601876747805505, 0.9522580699968279, 0.06304475092339623, 0.18909001622655808, 0.19934942931986965  …  0.021532597906190776, 0.8482825697641306, 0.10773487816863903, 0.32189982199036116, 0.12662208474317038, 0.28529465447429614, 0.2907506630258835, 0.36872799387588473, 0.061489791166806085, 0.45645058368583713], Column2 = [0.06546916714160167, 0.7243956502957003, 0.5183099801474415, 0.7555562860508294, 0.11226218114407538, 0.9135150277876691, 0.8739421974558176, 0.2268482788660101, 0.580604436651146, 0.4142252330250549  …  0.6517425913240111, 0.01713263102740481, 0.7175499403837856, 0.7362894157420817, 0.24893665902538054, 0.41499951381631595, 0.2159527717429719, 0.8966879835264249, 0.87252430655793, 0.41461921031276117], Column3 = [0.5939320702328891, 0.19329886972497456, 0.04656947038518311, 0.22095698685781184, 0.678807659662497, 0.12720198818430306, 0.6795750371448686, 0.9314917999820301, 0.22920734893984274, 0.5148148980955375  …  0.55049773593343, 0.038576459283091946, 0.27765727942909757, 0.2753072414696357, 0.8823620780359746, 0.44831794170895023, 0.9073846432163745, 0.4648550947905655, 0.311984726769037, 0.25829997798611304], Column4 = [0.12253944650540982, 0.8259140842535423, 0.4034477332184384, 0.5279399406265695, 0.5579944087437719, 0.24650366028608328, 0.6874897000162434, 0.23391406844015605, 0.5641254897013973, 0.6250622796341656  …  0.21708181942178983, 0.35224683896541464, 0.8444113778983325, 0.4547214584884428, 0.13508852017592232, 0.9510137735662383, 0.5723463533029658, 0.626377972762265, 0.7854013810594317, 0.15394691114473347], Column5 = [0.47958743625921163, 0.45779753417165514, 0.6367059235247621, 0.8601116026079643, 0.3334020182022719, 0.41593698717526373, 0.13208968772625174, 0.16951044109747648, 0.8137887839507706, 0.4429229861115882  …  0.01308976221980429, 0.48597926808091163, 0.20768781798463476, 0.30045611276046247, 0.15759293576302558, 0.975806377881983, 0.19451065500145392, 0.9638103356367584, 0.3594043445295293, 0.7792867217495332], Column6 = [3.0, 3.0, 1.0, 3.0, 1.0, 2.0, 3.0, 2.0, 3.0, 3.0  …  3.0, 2.0, 1.0, 2.0, 1.0, 2.0, 2.0, 3.0, 3.0, 1.0], Column7 = [2.0, 2.0, 2.0, 2.0, 1.0, 2.0, 2.0, 2.0, 1.0, 1.0  …  2.0, 1.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 1.0, 1.0]), CategoricalArrays.CategoricalValue{Int64, UInt32}[0, 0, 0, 0, 0, 0, 0, 0, 1, 0  …  0, 0, 1, 0, 1, 0, 0, 0, 0, 0])"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "X, y = generate_imbalanced_data(100, 5; cat_feats_num_vals = [3, 2], \n",
+    "                                        probs = [0.9, 0.1], \n",
+    "                                        type = \"ColTable\", \n",
+    "                                        rng=42)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Construct `BalancedBaggingClassifier` Model"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "BalancedBaggingClassifier(\n",
+       "  model = LogisticClassifier(\n",
+       "        lambda = 2.220446049250313e-16, \n",
+       "        gamma = 0.0, \n",
+       "        penalty = :l2, \n",
+       "        fit_intercept = true, \n",
+       "        penalize_intercept = false, \n",
+       "        scale_penalty_with_samples = true, \n",
+       "        solver = nothing), \n",
+       "  T = 10, \n",
+       "  rng = Xoshiro(0xa379de7eeeb2a4e8, 0x953dccb6b532b3af, 0xf597b8ff8cfd652a, 0xccd7337c571680d1))"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "LogisticClassifier = @load LogisticClassifier pkg=MLJLinearModels verbosity=0\n",
+    "logistic_model = LogisticClassifier()\n",
+    "model = BalancedBaggingClassifier(classifier=logistic_model, T=10, rng=Random.Xoshiro(42))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Train & Evaluate the Model"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "┌ Info: Training machine(LogisticClassifier(lambda = 2.220446049250313e-16, …), …).\n",
+      "└ @ MLJBase /Users/essam/.julia/packages/MLJBase/ByFwA/src/machines.jl:492\n",
+      "┌ Info: Solver: MLJLinearModels.LBFGS{Optim.Options{Float64, Nothing}, NamedTuple{(), Tuple{}}}\n",
+      "│   optim_options: Optim.Options{Float64, Nothing}\n",
+      "│   lbfgs_options: NamedTuple{(), Tuple{}} NamedTuple()\n",
+      "└ @ MLJLinearModels /Users/essam/.julia/packages/MLJLinearModels/zSQnL/src/mlj/interface.jl:72\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "100-element CategoricalDistributions.UnivariateFiniteVector{Multiclass{2}, Int64, UInt32, Float64}:\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.928, 1=>0.0722)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.845, 1=>0.155)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.749, 1=>0.251)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.902, 1=>0.0977)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.804, 1=>0.196)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.864, 1=>0.136)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.851, 1=>0.149)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.954, 1=>0.0458)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.853, 1=>0.147)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.86, 1=>0.14)\n",
+       " ⋮\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.671, 1=>0.329)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.73, 1=>0.27)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.843, 1=>0.157)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.941, 1=>0.0594)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.872, 1=>0.128)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.92, 1=>0.0797)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.929, 1=>0.0714)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.791, 1=>0.209)\n",
+       " UnivariateFinite{Multiclass{2}}(0=>0.827, 1=>0.173)"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "mach = machine(logistic_model, X, y)\n",
+    "fit!(mach)\n",
+    "pred = predict(mach, X)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Julia 1.8.5",
+   "language": "julia",
+   "name": "julia-1.8"
+  },
+  "language_info": {
+   "file_extension": ".jl",
+   "mimetype": "application/julia",
+   "name": "julia",
+   "version": "1.8.5"
+  },
+  "orig_nbformat": 4
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}