Developing forecast model of by-product for production planning case: UPM Jämsänkoski

In this thesis by-production of Jämsänkoski is studied in order to find out relationship be-tween order parameters to be able to see is there possibilities to improve forecasting. More accurate forecast would offer better performance both operational and sales wise. This is done by checking relations between order parameters and resulting by-production historical-ly using regression analysis and k-means to group order types regarding by-production. Methods are commonly used tools and used in similar purposes but to this specific case and field it is new approach. Regression offered very little relation to others but only one variable meaning forecasting is most accurate to do in a straightforward way focusing only to that parameter. Future studies in this case are not likely to offer more precis way to forecast but methods are valid regardless in other cases.