There are 5 companies that each own about 100 different sites in different cities. There is a new software system that is being distributed to each site at different times. The sites that have updated their system experience spikes in the number of issues at the time of update as well as some smaller spikes within the next couple of months. Each site is a different size (aka number of users added is different), each company has different experience with updating capability, and each update happens at a handful of sites per company when they opt to update their system. In July, there are a handful of sites for 3 of these companies that are updating their system.
My job is to analyze these past spike trends that occur with past updates in past sites to be able to predict how many issues the new sites that are about to update are going to have, so the companies can staff help accordingly.
I am struggling to fit a model or find the correct method that I need to use. Any advice?
I was thinking regression, but confused myself. if for instance I have the site chattanooga with total number of ticket and the properties of company 1, size medium, and experience 5 past sites active for the company, what would my regression look like? it is definitely nonlinear, but I am not sure how to incorporate the variables properly
right, I know that I can use all the variables to regress on number of issues, but I am still a little confused.
I understand how to do the regression using all sites that have updated and classifying them based on properties with my output as number of issues.
What i am wondering about is how to use one city to predict about another while still incorporating classifications/properties. This way I can predict which department the issues will be in for cities of the same size, same company, etc. Does my question make sense?
If you have other ideas on how to predict the issues per department, let me know!
I did linear models, and the fit isn't great. Should I try ridge regression/lasso? I am not very experienced with non linear regression.
I have fit a negative binomial regression model and now want to predict a future outcome wiht the knowledge of all the independent input variables, whcih are a mix of numerical and categorical variables.
Any advice on how to do this?
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers