List of questions

How can academia help us, the users of computational tools, when the tools come from other companies?

ABB Crane Systems optimizes transport of containers and steel products. They offer electrical and automation equipment for controlling the motions of container cranes, ship unloaders and industrial cranes. The goal is to achieve fast and cost-effective goods handling.
-------------------
Many companies active in traditional industrial segments are users of advanced tools for computations. These tools are developed by specialized companies and sold as products. Can academia help us, the users, directly or not? How?

A best estimate of the life expectancy is used for scoring endowment and pension insurance customers. Can we improve the precision of our customer scoring models?

Folksam is a Swedish insurance company that offers a wide variety of insurance, savings and loan products.
-------------
A best estimate of the life expectancy is used for scoring endowment and pension insurance customers. At present, the procedure most commonly used is based on multidimensional regression techniques (standard is using procedures available through the modeling packages provided by SAS institute). Can we improve the precision of our scoring models by applying alternative predictive modeling methods?

How can Geographical Information System (GIS) information contribute to improved prediction power of future claim costs and customer behaviour for insurance companies?

Folksam is a Swedish insurance company that offers a wide variety of insurance, savings and loan products.
-------------
The aim is to investigate the effect of using more detailed geographical information as descriptive variables in our prediction models for pricing and customer analytics.

How reliable is the data obtained from computer models compared with data generated by experiments on animals and in cell systems?

PROSILICO is a new company that develops new in-silico methods and visualization tools for drug development and analysis of environmental spreading of pharmaceuticals and chemicals.
----------------
Allt bättre datormodeller och visualiseringsverktyg för läkemedelsutveckling och miljöspridning av läkemedel och kemikalier utvecklas nu. Dock uppfattas det fortfarande som att data från djurförsök och data från cellsystem är mer robusta och tillförlitliga jämfört med data som tas fram med hjälp av datorberäkningar. Uppfattningen tycks vara att datorberäkningar ännu inte nått den resultatkvalitet som önskas och att dessa därför lämpar sig bäst för grov screening. Stämmer detta generellt eller finns det fall där datorberäkningar kan ersätta t.ex. djurförsök?

Is there a way to create a device whose spatial position can be described and the information be used in a CAD program?

Vegesack Arkitekter AB is a company of architects that work with construction, design, product development and decoration for buildings.
----------------
Vid uppmätning av byggnader finns idag ingen enkel och effektiv utrustning. Varje projektering kräver kännedom om byggnaderna på plats, noggranna planritningar i digitalt format, dessa underlag saknas ofta.
Teodoliter med en data enhet ibland kallad totalstation är ett långsamt och klumpigt instrument speciellt inomhus. Vår tanke är en mätpenna med vilken man registrerar punkter i rummet som direkt bildar en linje i cadprogrammet. En väldigt enkel idé egentligen. Mått-noggrannheten bör inte vara sämre än +-10 mm och det är också lämpligt att man inte behöver ha fri sikt vid mätningen, då detta komplicerar mätningarna i praktiken.

Bias in FHMMs (Factorial Hidden Markov Models): How should you handle the bias that occurs when using a FHMM, when there exist devices that are not modeled, but still contribute to the aggregate time series?

Watty develops a system which, from total energy data in a building, finds how much energy has gone to different component systems such as illumination, elevator, washing room etc. We collect total energy data from buildings every second and then try to classify how this energy has been used. Classification is done by training machine learning algorithms on data where the distribution of energy usage is known. The training data consist of large datasets with information about the total energy usage and its distribution in different systems. Our system could reduce the cost of energy declarations from ca 70000 kr to a few thousand kr.

Watty develops the first energy data analytics tool that can automate the energy efficiency process in buildings. Watty's patent-pending big data solution uses the total energy use in a building to evaluate its underlying systems and compare it to alternatives. Watty is a spinout from KIC Innoenergy.

In a time series, how do you handle a non-uniform sampling, missing data and event based sampling?  When resampling, how does it affect the behaviour of a FHMM (Factorial Hidden Markov Model)?                                                             

Watty develops a system which, from total energy data in a building, finds how much energy has gone to different component systems such as illumination, elevator, washing room etc. We collect total energy data from buildings every second and then try to classify how this energy has been used. Classification is done by training machine learning algorithms on data where the distribution of energy usage is known. The training data consist of large datasets with information about the total energy usage and its distribution in different systems. Our system could reduce the cost of energy declarations from ca 70000 kr to a few thousand kr.

Watty develops the first energy data analytics tool that can automate the energy efficiency process in buildings. Watty's patent-pending big data solution uses the total energy use in a building to evaluate its underlying systems and compare it to alternatives. Watty is a spinout from KIC Innoenergy.