文章目錄

• 0 Preface
• 1 載入資料
• 2 訓練隨機森林（train RF）
• 3 完整程式碼

0 Preface

相關引數說明

 - Julia: 1.0
 - OS: MacOS 

訓練測試資料百度雲連結：點選下載 密碼: u71o
檔案說明：

 - rf_julia_charReg
	 - resizeData.py    #批量重設定圖片尺寸
	 - test    #測試圖片檔案
	 - testResized    #resized 測試圖片檔案
	 - train    #訓練圖片檔案
	 - trainResized    #resized 訓練圖片檔案
	 - sampleTest.csv    #測試資料csv檔案
	 - trainLabels.csv     #訓練資料label csv檔案
 

1 載入資料

安裝需要使用到的包：

using Images
using DataFrames
using Statistics #use mean(), sum()... function
using DataFrames
using CSV

	注：如果沒有安裝包，使用以下指令碼安裝

import Pkg
Pkg.add([PKG NAME]) #例如：Pkg.add("Images")

讀取圖片檔案資料，並返回矩陣

function read_data(type_data, labelsInfo, imageSize, path)
    x = zeros(
 
size(labelsInfo, 1), imageSize)
    for (index, idImage) in enumerate(labelsInfo.ID)
        nameFile = "$(path)/$(type_data)Resized/$(idImage).Bmp"
	    img = load(nameFile)
        temp = float32(img)
        temp = Gray.(temp)
        x[index, :] = reshape(temp, 1, imageSize)
    end
    return x
end
 

解釋：

float32(): 將其中的值轉化為浮點數
Gray.(): 將RGB影象轉化為灰度影象
reshape(): 在這裡做的是平鋪工作

設定影象大小以及專案路徑：

imageSize = 400
path = "..."

讀取訓練資料Label

labelsInfoTrain = CSV.read("$(path)/trainLabels.csv")

讀取訓練影象資料：

xTrain = read_data("train", labelsInfoTrain, imageSize, path)

讀取測試資料Label:

labelsInfoTest = CSV.read("$(path)/sampleSubmission.csv")

讀取測試影象資料：

xTest = read_data("test", labelsInfoTest, imageSize, path)

2 訓練隨機森林（train RF）

訓練：

model = build_forest(yTrain, xTrain, 20, 50, 1.0)

解釋：

$3(20)：number of features chosen at each random split
$4(50): number of trees
$5(1.0): ratio of subsampling

獲得測試結果：

predTest = apply_forest(model, xTest)

轉化預測結果：

labelsInfoTest.Class = Char.(predTest)

寫入檔案：

CSV.write("$(path)/predTest.csv", labelsInfoTest, header=true)

四折交叉驗證：

accuracy = nfoldCV_forest(yTrain, xTrain, 20, 50, 4, 1.0);
println("4 fold accuracy: $(mean(accuracy))")

3 完整程式碼

using Images
using DataFrames
using Statistics 
using DataFrames
using CSV
using DecisionTree

function read_data(type_data, labelsInfo, imageSize, path)
    x = zeros(size(labelsInfo, 1), imageSize)
    for (index, idImage) in enumerate(labelsInfo.ID)
        nameFile = "$(path)/$(type_data)Resized/$(idImage).Bmp"
        img = load(nameFile)
        temp = float32(img)
        temp = Gray.(temp)
        x[index, :] = reshape(temp, 1, imageSize)
    end
    return x
end


imageSize = 400
path = "/Users/congying/cyWang/projects/julia/kaggleFirstStepsWithJulia/all"
labelsInfoTrain = CSV.read("$(path)/trainLabels.csv")
xTrain = read_data("train", labelsInfoTrain, imageSize, path)
labelsInfoTest = CSV.read("$(path)/sampleSubmission.csv")
xTest = read_data("test", labelsInfoTest, imageSize, path)
yTrain = map(x -> x[1], labelsInfoTrain.Class)
yTrain = Int.(yTrain)


model = build_forest(yTrain, xTrain, 20, 50, 1.0)
predTest = apply_forest(model, xTest)
labelsInfoTest.Class = Char.(predTest)
CSV.write("$(path)/juliaSubmission.csv", labelsInfoTest, header=true)
accuracy = nfoldCV_forest(yTrain, xTrain, 20, 50, 4, 1.0);
println("4 fold accuracy: $(mean(accuracy))")