[MSR] MSDN - Analyzing loan data with ScaleR(Logistic Regression)

/*******************************************************************************************************************
-- Title : [MSR] MSDN -  Analyzing loan data with ScaleR(Logistic Regression)
-- Reference : microsoft.com
-- Key word : microsoft r revoscale r scale r rximport rxsummary rxgetinfo  rxgetvarinfo rxlogit rxpredict
                  logistic regression 로지스틱 회귀분석 회귀 모델 summary 
*******************************************************************************************************************/

-- Chart

-- MSDN
    * https://msdn.microsoft.com/en-us/microsoft-r/scaler-getting-started-1-example-loan-data

-- 샘플 다운로드
    * http://packages.revolutionanalytics.com/datasets/

-- Microsoft R

# ------------------------------
# -- 샘플 다운로드 
# ------------------------------
#    http://packages.revolutionanalytics.com/datasets/
 
# ------------------------------
# -- 경로 설정
# ------------------------------
csvDataDir <- "C:\\RProject\\MRS\\Data"
mortCsvDataName <- file.path(csvDataDir, "mortDefault")                   # 파일네임 : mortDefault2000..csv, ...
 
mortXdfFileName <- "C:\\RProject\\MRS\\Data\\mortDefaultSmall.xdf"
 
 
# ------------------------------
# -- 여러 파일 가져오기(rxImport)
# ------------------------------
append <- "none"
for (i in 2000:2009) {
    importFile <- paste(mortCsvDataName, i, ".csv", sep = "")
    #cat(importFile)
 
    mortDS <- rxImport(importFile, mortXdfFileName, append = append)
    append <- "rows"                                                      # 기존 DS에 Append : "rows"
}
 
nrow(mortDS)
 
# ------------------------------
# -- DS 정보 확인(rxSummary, rxGetInfo, rxGetVarInfo)
# ------------------------------
 
# -- DS 기본 요약 및 5개의 행 확인
rxGetInfo(mortDS, numRows = 5)
 
# -- 필드 확인 
rxGetVarInfo(mortDS)
 
# -- 요약 통계 보기
rxSummary( ~ ., data = mortDS, blocksPerRead = 2)                         # DS 읽기 block 개수
rxSummary( ~ creditScore, data = mortDS, transforms = list(creditScore = cut(creditScore, breaks = 20)))
 
 
# ------------------------------
# -- 로지스틱 회귀 계산(rxLogit)
# ------------------------------
 
# -- 기본 Logistic Regression
logitObj <- rxLogit(default ~ F(year) + creditScore +
            yearsEmploy + ccDebt,
            data = mortDS, blocksPerRead = 2,
            reportProgress = 1)
 
summary(logitObj)
 
# -- 매개변수가 많은 Logistic Regression
system.time(
            logitObj <- rxLogit(default ~ F(houseAge) + F(year) +
                        creditScore + yearsEmploy + ccDebt,
                        data = mortDS, blocksPerRead = 2, reportProgress = 1))
 
summary(logitObj)
 
# -- 플롯팅
cc <- coef(logitObj)
df <- data.frame(Coefficient = cc[2:41], HouseAge = 0:39)
rxLinePlot(Coefficient ~ HouseAge, data = df, type = "p")
 
# ------------------------------
# -- 기본 확률값 계산(rxPredict)
# ------------------------------
 
# -- 관심 벡터 정의
creditScore <- c(300, 700)
yearsEmploy <- c(2, 8)
ccDebt <- c(5000, 10000)
year <- c(2008, 2009)
houseAge <- c(5, 20)
 
# -- 데이터 프레임 생성
predictDF <- data.frame(
        creditScore = rep(creditScore, times = 16),
        yearsEmploy = rep(rep(yearsEmploy, each = 2), times = 8),
        ccDebt = rep(rep(ccDebt, each = 4), times = 4),
        year = rep(rep(year, each = 8), times = 2),
        houseAge = rep(houseAge, each = 16))
 
head(predictDF)
 
predictDF2 <- rxPredict(modelObject = logitObj, data = predictDF,
             outData = predictDF)
 
predictDF2[order(predictDF2$default_Pred, decreasing = TRUE),]