Should we expect deteriorated performance while applying a learned model compared with fit command ?

[bibinksebastian]

I am trying to forecast Tput_Gbps using smartforecasting . It works well when I fit the model . SPL: source="150daysABC - Copy.csv" host="forecast" index="150dayscopy" sourcetype="csv"
| reverse
| table _time Tput_Gbps Holiday_Flag | fit StateSpaceForecast "Tput_Gbps" as smartforecast output_metadata=true holdback=0 forecast_k=384 conf_interval=95 period=96 specialdays=Holiday_Flag into "_150_CD"
| tail 500 | table _time smartforecast Tput_Gbps
![alt text][1]

However, on the same data I tried to apply the model trained earlier. SPL: source="150daysABC - Copy.csv" host="forecast" index="150dayscopy" sourcetype="csv"
| reverse
| table _time Tput_Gbps Holiday_Flag | apply _150_CD

| tail 500 | table _time predicted(Tput_Gbps) Tput_Gbps

![alt text][2]

Towards the end of the prediction, I am seeing multiple predictions too. Looks very strange !! not sure how and why . Spent hours on troubleshooting this.

Please help to check if it is a bug or something that I am doing which is not right. @grana_splunk .This is in continuation to my earlier question. Thanks in advance for the help

[grana_splunk]

Is there any possibility, I can see your sample data? Without looking into it, I am not able to debug this issue further.

[bibinksebastian]

I tried without specialdays input and specifying all the parameter necessary for apply method.
source="150daysAB - Copy.csv" host="forecast" index="150dayscopy" sourcetype="csv"
| reverse
| table _time Tput_Gbps Holiday_Flag | apply _150_AB3 holdback=0 forecast_k=384 conf_interval=95 period=96

| tail 500 | table _time predicted(Tput_Gbps) Tput_Gbps. It still gives me the same bad output