Time Series forecast on multiple labels see sample data

sunnyal

Hi,

I would like to forecast the attached data for future months for each client name. How can I do this? I'm comfortable to forecast for single client but how can we achieve for all clients in a single process. Please see attached sample data format.

Thx for your guidance

lionelderkrikor

Hi@sunnyal,

If you want to use a model per customer, you have to build one process per customer (in your case 5).

You can also useLoop Attributesto forecast the evolution of all your customers. In this case, you can only use the same model for all your attributes (your customers). In your case, I thinkkNNmodel is the good choice.

But I have diificulties to recover the timestamp after performing the forecasting and the "merging" of your different attributes.

Anyways, you can find the process here :

<参数键= =”可能17. true.integer.att“7”价值ribute"/>

















































<参数键= =”可能17. true.integer.att“7”价值ribute"/>


















































<参数键= value =“attribute_name % {loop_attribute}"/>




















<参数键= value =“attribute_name % {loop_attribute}"/>







<运营商激活= " true " class = "循环”compatibility="9.0.000-BETA4" expanded="true" height="82" name="Loop" width="90" x="782" y="340">







































































































<连接from_op = "日期转换”from_port = "考试ple set output" to_op="Multiply" to_port="input"/>

























Data loading and preparation
Extends weeks column to the length of horizon period in the future. 'Number of interations' parameter in loop block should be equal to horizon parameter of 'predict series' block


















< portSpacing端口= " sink_result 3”温泉cing="0"/>

You have to install the "Series Extension" extension (from the marketplace) to have thePredict Seriesoperator.

I hope it helps,

Regards,

Lionel

JEdward

You can use the LoopValues operator and store the result in a macro to build a model for each customer.

This is a rough example of this, in a real case you would be doing a lot more preprocessing and sometimes storing results to disk for later processing.

sunnyal

@JEdward@lionelderkrikorThank you for the response.

@JEdwardHow do you define a label here, on which attribute?? I tried your process and it complains about missing label??

sunnyal

@JEdwardOk I got the model working, when I set role and define a label. However, the store oprator stores the model output for each client, so now how do I forecast this forall clients at oncefor future dates (i.e.) till December.

Is there a operator to join all the model outpus ??