Example 41.1. Building time series model

tableName="stock_market"
targetName="price_change"
timeAttrName="time"


if not tableExists(tableName):
 raise "Table '"+tableName+"' does not exists. Please run "+tableName+".py script from data directory first" 



pd = PhysicalData(tableName)
ld = LogicalData(pd)
save('ts_pd', pd)
save('ts_ld', ld)


fs = TimeSeriesFunctionSettings()

#----- GARCH algorithm settings
as = GARCHSettings()
as.executeInitTests = TRUE
as.groupStatistics = FALSE
as.preselection = TRUE

opt = OptimizationAlgorithmSettings()
opt.convergenceCriterion = ConvergenceCriterion.likelihood
opt.convergenceThreshold = 0.001
opt.iterMax = 100

vss = VariableSelectionSettings()
vss.variableSelectionMethod = VariableSelectionMethod.full
vss.modelEntryLevel = 0.15
vss.modelLeaveLevel = 0.2
vss.groupMode = FALSE

as.variableSelectionSettings = vss
as.optimizationAlgorithmSettings = opt
fs.algorithmSettings = as


#----- time series function settings
fs.logicalData = ld
fs.targetAttributeName = targetName
fs.timePointAttributeName = timeAttrName

#fs.attributeUsageSet.getAttribute('season').setUsage(UsageOption.inactive)
#fs.attributeUsageSet.getAttribute('seeded').setUsage(UsageOption.inactive)


save('ts_settings', fs)


#----- building time series model
bt = MiningBuildTask('ts_pd', 'ts_settings', 'ts_model')

save('ts_build', bt)
execute('ts_build')


# print model statistics
statModel = load('ts_model').getModelStatistics()
varStat = statModel.variableStatistics
statNames = varStat.varStatNames
varNames=statModel.varNames

print " "*5,
for colName in statNames:
    print "%10s"%colName[0:10],
print

for row in range(0,len(varNames)):
        print "%5s"%varNames[row][0:20],
        for col in range(0,len(statNames)):
            value = (str(varStat.getVarStatValue(varNames[row],col)))
            if value == "NaN": value = "%10s" % value
            else: value = "%10.4f" % value
            print value,
        print

Output:

            Coeff Univariate Lower Conf Upper Conf     StdErr  Wald Test Wald Pr>Ch
alpha_0    -0.0000     0.0000    -0.0000    -0.0000     0.0000 14661.3023     0.0000
alpha_1     0.2157     0.0000     0.2154     0.2160     0.0002 1889852.0919     0.0000
beta_1     0.0653     0.0000     0.0652     0.0654     0.0000 2381730.0596     0.0000
Intercept    -0.0001     0.0000    -0.0001    -0.0001     0.0000 72473.1732     0.0000

Example 41.2. Time series model testing

tsTestTask = TimeSeriesTestTask()
tsTestTask.setTestResultName("tsTestResult")
tsTestTask.setModelName("ts_model") 
tsTestTask.setTestDataName("ts_pd")
tsTestTask.setTestDataTargetAttributeName("price_change")

save('tsTestTask',tsTestTask)
execute('tsTestTask')

model = load('tsTestResult')

print "Absolute error:", "%f" % model.meanAbsoluteError
print "Actual value:", "%f" % model.meanActualValue
print "Predicted variance:", "%f" % model.meanPredictedVariance
   

Output:

Absolute error: 0.000335
Actual value: -0.000311
Predicted variance: 0.000077

Example 41.3. Time series model application

apply_pd = PhysicalData('stock_market_res')
save('stock_market_res_pd', apply_pd )

tSerAppTask = MiningApplyTask()
tSerAppTask.modelName = 'ts_model'
tSerAppTask.sourceDataName = 'ts_pd'
tSerAppTask.targetDataName = 'stock_market_res_pd'
tSerAppTask.replaceExistingData = TRUE

directMapping = java.util.ArrayList()
directMapping.add( ApplySourceItem('price_change','price_change') )
directMapping.add( ApplySourceItem('time','time') )
tSerAppTask.setDirectMapping(directMapping)

tSerAppOutput = TimeSeriesApplyOutput()


tSerAppTask.applyOutput = tSerAppOutput

save('tSer_apply_task',tSerAppTask)
execute('tSer_apply_task')

print "Print first 10 rows"
print "%-15s%-13s%-13s%-21s%-21s%-13s" % ('price_change', 'time', 'variance__1', 'forecast_lowerbound','forecast_upperbound', 'regressed_mean')
strFormat = "%-15.1f%-13.1d%-13.2f%-21.5s%-21.5s%-13.2f"
counter = 0
trans None<-'stock_market_res':
    $counter+=1
    if $counter>=10: __exit__=1
    print $strFormat % (price_change, time, variance__1, $str(forecast_lowerbound),$str(forecast_upperbound), regressed_mean)
    

Output:

Print first 10 rows
price_change   time         variance__1  forecast_lowerbound  forecast_upperbound  regressed_mean
0.0            1            0.00         -0.01                0.018                -0.00        
-0.0           2            0.00         -0.00                0.004                -0.00        
-0.0           3            0.00         -0.01                0.013                -0.00        
-0.0           4            0.00         -0.00                0.006                -0.00        
0.0            5            0.00         -0.00                0.002                -0.00        
-0.0           6            0.00         -0.00                0.007                -0.00        
-0.0           7            0.00         -0.00                0.002                -0.00        
-0.0           8            0.00         -0.00                0.008                -0.00        
0.0            9            0.00         -0.02                0.023                -0.00        
0.0            10           0.00         -0.00                0.009                -0.00