Convert Advance Analytics Tool Box Expression to KPI Expression

[phaneendra_kunc]

HI,

Can any one help me to work below expression in KPI object? Below one is captured from "Console window in Chrome" which was produced from a "Analytics Tool Box extension"

R.ScriptEvalExStr('N', 'save(q,file="~/debug_holt_winters.rda"); library(jsonlite);library(dplyr);library(forecast);data<-ts(na.omit(q$Measure) ,frequency=12);

      fit<-HoltWinters(data, seasonal="additive" );

      res<-forecast(fit, level=0.95, h=15);

      json<-toJSON(list(as.double(res$mean),as.double(res$upper),as.double(res$lower),list(fit$alpha, fit$beta, fit$gamma, fit$coefficients))); json;', Sum(Sales) as Measure)

This works good in the Advanced Analytics Toolbox --> Time Series --> HoltWinters

Thanks,

Phani

[Anonymous]

Hi phaneendra.kunche‌!

Without looking further into the R-script itself my guess would be that you're trying to use a tensor function (since you're using ScriptEvalExStr) instead of an aggregation. KPI objects only allow a single value to be returned. Does your function return multiple values? If so, try using a line chart or similar.

Kind regards,

Josefine

[phaneendra_kunc]

Thanks for the reply.

You are right, above expression returns multiple values. I'm looking to convert to a KPI expression where it results sum of forecast results for 15 future months.

Example: I have 2 years historical sales data by Year Month and using Holt Winters model, i want to plot the data data in a line chart, which i was able to do that using "Analytics Tool Box" extension. But i need to show the sum of forecast results in a KPi object. I don't have to use the Analytics toolbox, just need an expression where it AGGREGATES using HOLT WINTERS model.

Installed R Packages:

Rserve,ChainLadder,colorspace,caret,jsonlite,tseries,forecast,tm,dplyr,SnowballC,cluster,rpart,d3r,partykit

Sample QVF:

Attached

Any  help would be greatly appreciated.

Thanks,

Phani

[Anonymous]

Hi again!

I'm not sure exactly what you want to sum if it's the mean, upper, lower or other. But what I would do at first is to change the function to ScriptAggr and second, remove the json part (which I think is only needed for the "Analytics Tool Box" extension). What you want there instead is the sum of your choice, as an example, I used "sum(res$mean)", which you should change according to your preference. Unfortunately, I haven't been able to test the code, but hopefully, this will help you going forward.

R.ScriptAggr('save(q,file="~/debug_holt_winters.rda"); library(jsonlite);library(dplyr);library(forecast);data<-ts(na.omit(q$Measure) ,frequency=12);

      fit<-HoltWinters(data, seasonal="additive" );

      res<-forecast(fit, level=0.95, h=15);

      sum(res$mean);', Sum(Sales) as Measure)

Regards,

Josefine

[phaneendra_kunc]

I tried your expression, but still not resulting into any value.

[yannick_c]

Hi Phaneendra

I have a solution : replacing ScriptAggr with ScriptEval and aggr()

aggr(

R.ScriptEval('

library(forecast);

data<-ts(na.omit(q$Measure) ,frequency=12);

fit<-HoltWinters(data, seasonal="additive" );

res<-forecast(fit, level=0.95, h=15);

sum(res$mean);',

Sum(Sales) as Measure),

YearMonth)

Regards