public static void createDataFrame() {
// Convert batch to JSON and inject into R native dataframe
// This will be required for JSON processing
System.out.println(engine.eval("library(jsonlite)"));
for (int i = 0; i < hdfsPaths.size(); i++) {
System.out.println(
engine.eval(
"testdata_" + i + " <- read.df(sqlContext, '" + hdfsPaths.get(i) + "', 'json')"));
}
}
public static void readRScript(String pathToRScript) {
Path file = Paths.get(pathToRScript);
try (Stream<String> lines = Files.lines(file, Charset.defaultCharset())) {
lines.forEachOrdered(line -> System.out.println(engine.eval(line)));
} catch (IOException e) {
e.printStackTrace();
}
}
private static REXP runCommand(String command, boolean convert) {
boolean obtainedLock = engine.getRsync().safeLock();
REXP value;
try {
value = engine.eval(command, convert);
} finally {
if (obtainedLock) engine.getRsync().unlock();
}
if (!convert) engine.waitForR();
return value;
}
public static void runTestScript() {
System.out.println(
engine.eval(
"jrdd <- SparkR:::callJStatic(\"com.ikanow.aleph2.analytics.r.utils.RScriptUtils\",\"getJavaRDD\", sc)"));
System.out.println(engine.eval("SparkR:::show(jrdd)"));
System.out.println(engine.eval("newRdd <- SparkR:::RDD(jrdd, \"string\")"));
System.out.println(engine.eval("df <- SparkR:::createDataFrame(sqlContext, newRdd)"));
System.out.println(engine.eval("showDF(df)"));
System.out.println(engine.eval("localDf <- collect(df)"));
System.out.println(engine.eval("localDf"));
}
public static void start(String... args) {
if (!Rengine.versionCheck()) {
System.err.println("** Version mismatch " + "- Java files don't match library version. **");
System.exit(1);
}
System.out.println("Creating R Engine (with arguments)");
// 1) we pass the arguments from the command line
// 2) we won't use the main loop at first, we'll start it later
// (that's the "false" as second argument)
// 3) the callbacks are implemented by the TextConsole class above
engine = new Rengine(args, false, null);
// the engine creates R is a new thread, so we should wait until it's ready
if (!engine.waitForR()) {
System.out.println("Cannot load R.");
return;
}
engine.eval("{library(rJava);.jinit()}", false);
}
public static void main(String[] args) throws IOException {
start("--vanilla");
initializeSparkR();
System.out.println(
engine.eval(
"jrdd <- SparkR:::callJStatic(\"com.ikanow.aleph2.analytics.r.utils.RScriptUtils\",\"getJavaRDD\", sc)"));
System.out.println(engine.eval("SparkR:::show(jrdd)"));
System.out.println(engine.eval("newRdd <- SparkR:::RDD(jrdd, \"string\")"));
System.out.println(engine.eval("df <- SparkR:::createDataFrame(sqlContext, newRdd)"));
System.out.println(engine.eval("showDF(df)"));
System.out.println(engine.eval("localDf <- collect(df)"));
System.out.println(engine.eval("localDf"));
// processAnalytics();
// readRScript();
shutdown();
}
public static void initializeSparkR() {
start("--vanilla");
// To do move these into an array or separate file or something and check for null
// if null we need to exit our gracefully as all statements are required for the jobs to run
System.out.println(engine.eval(".libPaths(c(.libPaths(), '/root/spark-1.5.2/R/lib'))"));
System.out.println(engine.eval("Sys.setenv(SPARK_HOME = '/root/spark-1.5.2')"));
System.out.println(
engine.eval(
"Sys.setenv(PATH = paste(Sys.getenv(c('PATH')), '/root/spark-1.5.2/bin', sep=':'))"));
System.out.println(engine.eval("library(SparkR)"));
// We should probably point to the server here Im guessing this is pointing to localhost
System.out.println(
engine.eval(
"sc <- sparkR.init(sparkJars = \"/root/aleph2_analytic_services_R-0.0.1-SNAPSHOT.jar\")"));
System.out.println(engine.eval("sqlContext <- sparkRSQL.init(sc)"));
}
// Run prediction method on data based on file name
public StockInfo[] runPrediction(Map<String, StockInfo[]> dataToProcess) {
double[] resultArrayOpen = null;
double[] resultArrayHigh = null;
double[] resultArrayLow = null;
double[] resultArrayClose = null;
double[] resultArrayVolume = null;
StockInfo[] predictedDataArray = null;
try {
logger.log("prediction will start");
// Flat the data to perform prediction on it
StockInfo[] flatStockInfo = this.flatMapOfData(dataToProcess);
Collections.reverse(Arrays.asList(flatStockInfo));
REXP x;
RVector v;
// Check for installed packages
x = re.eval("installed.packages()");
v = x.asVector();
String[] packages = x.asStringArray();
boolean isForecastInstalled = false;
logger.log("<R> getting installed packages");
for (int index = 0; index < packages.length && isForecastInstalled == false; index++) {
logger.log("<R> has installed " + packages[index]);
if (packages[index] != null && packages[index].compareTo("forecast") == 0) {
isForecastInstalled = true;
}
}
// If forecast needs to be installed
if (isForecastInstalled == false) {
logger.log("<R> will set repos");
// Set CRAN
re.eval("r <- getOption(\"repos\")");
re.eval("r[\"CRAN\"] <- \"http://cran.us.r-project.org\"");
re.eval("options(repos = r)");
re.eval("rm(r)");
// Install forecast
re.eval("install.packages(\"forecast\")");
logger.log("<R> will install forecast package");
}
// Load forecast library
re.eval("library(\"forecast\")");
logger.log("<R> loaded forecast");
// Make prediction for Open value -----------------------
// Load data into R
logger.log("<R> loading data into R");
StringBuilder builder = new StringBuilder("inputData <- c(");
for (int index = 0; index < flatStockInfo.length; index++) {
builder.append(flatStockInfo[index].open);
if (index != flatStockInfo.length - 1) {
builder.append(",");
} else {
builder.append(")");
}
}
String stringFromBuilder = builder.toString();
re.eval(stringFromBuilder);
// Create time series from data
logger.log("<R> forecasting open values BestFit");
re.eval("temporalData <- ts(inputData, frequency=365)");
// Forecast data
re.eval("forecastData <- forecast(temporalData, h=30)");
// re.eval("arimaModel <- auto.arima(temporalData, max.p=5, max.q=5, max.P=5, max.Q=5)");
// re.eval("forecastData <- forecast(arimaModel, h=30)");
x = re.eval("forecastData");
v = x.asVector();
x = (REXP) v.elementAt(1); // instead of 3
resultArrayOpen = x.asDoubleArray();
// Make prediction for High value ------------------------
builder = new StringBuilder("inputData <- c(");
for (int index = 0; index < flatStockInfo.length; index++) {
builder.append(flatStockInfo[index].high);
if (index != flatStockInfo.length - 1) {
builder.append(",");
} else {
builder.append(")");
}
}
stringFromBuilder = builder.toString();
re.eval(stringFromBuilder);
// Create time series from data
logger.log("<R> forecasting high values BestFit");
re.eval("temporalData <- ts(inputData, frequency=365)");
// Forecast data
re.eval("forecastData <- forecast(temporalData, h=30)");
// re.eval("arimaModel <- auto.arima(temporalData, max.p=5, max.q=5, max.P=5, max.Q=5)");
// re.eval("forecastData <- forecast(arimaModel, h=30)");
x = re.eval("forecastData");
v = x.asVector();
x = (REXP) v.elementAt(1);
resultArrayHigh = x.asDoubleArray();
// Make prediction for Low value ------------------------
builder = new StringBuilder("inputData <- c(");
for (int index = 0; index < flatStockInfo.length; index++) {
builder.append(flatStockInfo[index].low);
if (index != flatStockInfo.length - 1) {
builder.append(",");
} else {
builder.append(")");
}
}
stringFromBuilder = builder.toString();
re.eval(stringFromBuilder);
// Create time series from data
logger.log("<R> forecasting low values BestFit");
re.eval("temporalData <- ts(inputData, frequency=365)");
// Forecast data
re.eval("forecastData <- forecast(temporalData, h=30)");
// re.eval("arimaModel <- auto.arima(temporalData, max.p=5, max.q=5, max.P=5, max.Q=5)");
// re.eval("forecastData <- forecast(arimaModel, h=30)");
x = re.eval("forecastData");
v = x.asVector();
x = (REXP) v.elementAt(1);
resultArrayLow = x.asDoubleArray();
// Make prediction for Close value ------------------------
builder = new StringBuilder("inputData <- c(");
for (int index = 0; index < flatStockInfo.length; index++) {
builder.append(flatStockInfo[index].close);
if (index != flatStockInfo.length - 1) {
builder.append(",");
} else {
builder.append(")");
}
}
stringFromBuilder = builder.toString();
re.eval(stringFromBuilder);
// Create time series from data
logger.log("<R> forecasting close values BestFit");
re.eval("temporalData <- ts(inputData, frequency=365)");
// Forecast data
re.eval("forecastData <- forecast(temporalData, h=30)");
// re.eval("arimaModel <- auto.arima(temporalData, max.p=5, max.q=5, max.P=5, max.Q=5)");
// re.eval("forecastData <- forecast(arimaModel, h=30)");
x = re.eval("forecastData");
v = x.asVector();
x = (REXP) v.elementAt(1);
resultArrayClose = x.asDoubleArray();
// Make prediction for Close value ------------------------
builder = new StringBuilder("inputData <- c(");
for (int index = 0; index < flatStockInfo.length; index++) {
builder.append(flatStockInfo[index].volume);
if (index != flatStockInfo.length - 1) {
builder.append(",");
} else {
builder.append(")");
}
}
stringFromBuilder = builder.toString();
re.eval(stringFromBuilder);
// Create time series from data
re.eval("temporalData <- ts(inputData, frequency=365)");
// Forecast data
re.eval("forecastData <- forecast(temporalData, h=30)");
// re.eval("arimaModel <- auto.arima(temporalData, max.p=5, max.q=5, max.P=5, max.Q=5)");
// re.eval("forecastData <- forecast(arimaModel, h=30)");
x = re.eval("forecastData");
v = x.asVector();
x = (REXP) v.elementAt(1);
resultArrayVolume = x.asDoubleArray();
// Create a single StockInfo[] for all data
StockInfo predictedData;
predictedDataArray = new StockInfo[30];
Date lastDate = flatStockInfo[flatStockInfo.length - 1].date;
Calendar c = Calendar.getInstance();
c.setTime(lastDate);
c.add(Calendar.DATE, 1);
logger.log("<R> values for forecasted data");
SimpleDateFormat dateFormat = new SimpleDateFormat();
dateFormat.applyPattern("dd/MM/YYYY");
// For all days that were predicted
for (int index = 0; index < 30; index++) {
predictedData = new StockInfo();
predictedData.open = (float) resultArrayOpen[index];
float maxHigh =
(float)
StrictMath.max(
resultArrayClose[index],
StrictMath.max(resultArrayHigh[index], resultArrayOpen[index]));
predictedData.high = maxHigh;
float minLow =
(float)
StrictMath.min(
resultArrayClose[index],
StrictMath.min(resultArrayLow[index], resultArrayOpen[index]));
predictedData.low = minLow;
predictedData.close = (float) resultArrayClose[index];
predictedData.volume = (int) resultArrayVolume[index];
while (c.get(Calendar.DAY_OF_WEEK) == Calendar.SUNDAY
|| c.get(Calendar.DAY_OF_WEEK) == Calendar.SATURDAY) {
c.add(Calendar.DATE, 1);
}
predictedData.date = c.getTime();
predictedDataArray[index] = predictedData;
logger.log(
"<R> stock prediction "
+ dateFormat.format(predictedData.date.getTime())
+ " open: "
+ predictedData.open
+ " high: "
+ predictedData.high
+ " low: "
+ predictedData.low
+ " close: "
+ predictedData.close);
c.add(Calendar.DATE, 1);
}
} catch (Exception e) {
logger.logException(e);
}
return predictedDataArray;
}
