julia_029_hw.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Julia 機器學習：LIBSVM 支撐向量機\n",
    "\n",
    "## 作業 029：\n",
    "\n",
    "請使用 SVM 來建立一個預測模型，來對手寫數字辨識資料集，建立一個分類模型，預測數字為何。\n",
    "\n",
    "本作業需要使用到的套件有 MLDatasets，請在執行以下範例前先安裝。\n",
    "\n",
    "```\n",
    "] add MLDatasets\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "using LIBSVM, MLDatasets, Statistics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 讀取資料"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "呼叫 MNIST 資料集的過程中，會先去檢查以前是否有下載過，如果是第一次下載，過程中會詢問是否下載資料集，請回答 `y`。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(FixedPointNumbers.Normed{UInt8,8}[0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; … ; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8]\n",
       "\n",
       "FixedPointNumbers.Normed{UInt8,8}[0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; … ; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8]\n",
       "\n",
       "FixedPointNumbers.Normed{UInt8,8}[0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; … ; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8]\n",
       "\n",
       "...\n",
       "\n",
       "FixedPointNumbers.Normed{UInt8,8}[0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; … ; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8]\n",
       "\n",
       "FixedPointNumbers.Normed{UInt8,8}[0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; … ; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8]\n",
       "\n",
       "FixedPointNumbers.Normed{UInt8,8}[0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; … ; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8], [7, 2, 1, 0, 4, 1, 4, 9, 5, 9  …  2, 3, 9, 0, 1, 2, 2, 0, 8, 9])"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_X, train_y = MNIST.traindata(1:10000)\n",
    "test_X,  test_y  = MNIST.testdata(1:1000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 資料轉換"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1000×784 Array{N0f8,2} with eltype FixedPointNumbers.Normed{UInt8,8}:\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0  …  0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0  …  0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0  …  0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " ⋮                        ⋮              ⋱                 ⋮              \n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0  …  0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0  …  0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0\n",
       " 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0     0.0  0.0  0.0  0.0  0.0  0.0  0.0"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_X = Matrix(reshape(train_X, 28*28, :)')\n",
    "test_X = Matrix(reshape(test_X, 28*28, :)')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_y = [\"$x\" for x in train_y]\n",
    "test_y = [\"$x\" for x in test_y];"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 預測"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SVC(LIBSVM.Kernel.RadialBasis, 0.0012755102040816326, nothing, 1.0, 3, 0.0, 0.001, true, false, false, LIBSVM.SVM{String}(SVC, LIBSVM.Kernel.RadialBasis, nothing, 784, 10, [\"5\", \"0\", \"4\", \"1\", \"9\", \"2\", \"3\", \"6\", \"7\", \"8\"], Int32[1, 2, 3, 4, 5, 6, 7, 8, 9, 10], Float64[], Int32[], LIBSVM.SupportVectors{String,FixedPointNumbers.Normed{UInt8,8}}(5196, Int32[649, 327, 536, 368, 668, 522, 595, 416, 517, 598], [\"5\", \"5\", \"5\", \"5\", \"5\", \"5\", \"5\", \"5\", \"5\", \"5\"  …  \"8\", \"8\", \"8\", \"8\", \"8\", \"8\", \"8\", \"8\", \"8\", \"8\"], FixedPointNumbers.Normed{UInt8,8}[0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; … ; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8; 0.0N0f8 0.0N0f8 … 0.0N0f8 0.0N0f8], Int32[1, 12, 36, 66, 101, 133, 139, 174, 176, 183  …  9926, 9937, 9938, 9945, 9949, 9953, 9957, 9971, 9983, 9997], LIBSVM.SVMNode[LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0)  …  LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0), LIBSVM.SVMNode(1, 0.0)]), 0.0, [0.0 0.0 … 1.0 0.0; 0.0 0.0 … 0.0 1.0; … ; -0.0 -0.0 … -0.0 -0.0; -0.0 -0.0 … -0.0 -0.0], Float64[], Float64[], [-0.22824513560570694, -0.8368986068412834, -1.1622220524866105, -0.8673260023376893, 0.21943747220376308, -0.3298503371613154, -0.5699379699514623, -0.5323086714596027, -0.7843797821858961, -0.34859621336132457  …  -0.3590618101043141, -0.7330689482475496, -0.6896694013633632, -0.7349715094439786, -0.5106496806863642, -0.4114127518423237, -0.5750126766335187, 0.017161626070966933, 0.28097815377530766, 0.3778254533553624], 3, 0.0012755102040816326, 200.0, 0.001, 1.0, 0.5, 0.1, true, false))"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg=LIBSVM.fit!( SVC(), train_X, train_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "ŷ = String[]\n",
    "for i = 1:size(test_X, 1)\n",
    "    y_ = LIBSVM.predict(reg, Matrix(test_X[i, :]'))\n",
    "    append!(ŷ, y_)\n",
    "end"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 評估"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "89.9"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy() = mean((ŷ .== test_y))*100\n",
    "accuracy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Julia 1.5.0",
   "language": "julia",
   "name": "julia-1.5"
  },
  "language_info": {
   "file_extension": ".jl",
   "mimetype": "application/julia",
   "name": "julia",
   "version": "1.5.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}