Regression model to predict resolve time gives very poor result
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-16-2023 01:37 PM
Anyone aware of which regression algorithms are used to predict the resolve time of an incident ?

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-17-2023 09:29 AM
Hi Adeb, currently XGboost is used as the regression algo. Have you gone through some of the tuning recommendations for regression?
Also did you assess the data quality of your training set?
-Lener
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-17-2023 01:06 PM
So we are using short desc to predict the resolve time of an incident.
We used TFIDF to vectorize the text after doing preprocessing but that does not help.
Target -Resolve time (in sec) ranges from approx 20 mins to 10 days.
I also tried with building the regression model in python using same data .
Performed standard preprocessing steps to cleanse, used tfidf vect and used regression algorithms like linear ,dtree,xgboost but accuracy is only around 28-30%.
Is there any series for regression to estimate resolve time similar to https://www.youtube.com/watch?v=sKIoumkelnI