ARIMA convert integer machine learning


I'm trying to run this query below:

(index=A sourcetype=jobs_info JOB_NAME IN (ACQUA)) OR (index=B sourcetype=FIRE) OR (index=C sourcetype=EARTH)

| eval _time = strftime(_time, "%Y-%m-%d")
| eval START_TIME = strptime(START_TIME,"%Y%m%d%H%M%S")
| eval END_TIME = strptime(END_TIME,"%Y%m%d%H%M%S")

| eventstats avg(EXECUTION_TIME) as avg stdev(EXECUTION_TIME) as stdev

| eval lowerBound=(avg-stdev*exact(1.5)), upperBound=(avg+stdev*exact(1.5))
| eval isOutlier=if(EXECUTION_TIME < lowerBound OR EXECUTION_TIME > upperBound, 1, 0)

| stats values(EXECUTION_TIME) as EXECUTION_TIME sum(TNeg) as neg by _time
| where isnotnull(EXECUTION_TIME)
| table _time neg EXECUTION_TIME
| sort - _time

| fit RandomForestRegressor EXECUTION_TIME from "_time" "neg" n_estimators=15 into "teste"
| apply "teste"
| eval predicted(EXECUTION_TIME) = round('predicted(EXECUTION_TIME)', 2)

| stats values(neg) as neg, values(EXECUTION_TIME) as REALEXEC, values(predicted(EXECUTION_TIME)) as EXEC by _time
| eval erro = round(((EXEC/REALEXEC)-1)*100, 2)
| eval _time = tonumber(_time)
| table _time neg REALEXEC EXEC
| sort _time
| fit ARIMA _time EXEC holdback=3 conf_interval=95 order=12-0-1 forecast_k=5 as prediction | forecastviz(5, 3, "EXEC", 95)

And I'm having this error: Error in 'fit' command: Error while fitting "ARIMA" model: cannot convert float NaN to integer.
How can I can fix it and is there some easier way to run my code?

0 Karma

Splunk Employee
Splunk Employee

Either your _time or EXEC could be in float format which needs to be changed to the integer type.
Could you show the table for _time and EXEC just before the fit ARIMA command?

0 Karma