Marcus Skyum Email & Phone Number

Munich, By, De

Amsterdam, North Holland, Netherlands

Develop advanced analytical methods to aid the platform work at Systemiq.Ranging from complex and dynamic modelling approaches, agent-based modelling, and geospatial models. Create interactive visualisations and actionable model output.

Copenhagen, Capital Region, Denmark

Developing advanced statistical models using machine and deep learning techniques for two research projects with the aim of contributing to novel findings public health research. Project 1 concerns a model-based approach to estimate the risk of being acutely readmitted to the hospital within 6 months of the most recent discharge. This model can prove not only relevant in the healthcare sector, but also presents a method for calibrating societal health in public health studies. The study relies on 20 years of electronic medical records of adolescents and children in Denmark and uses a Convolutional Neural Network to identify diagnoses and predict the likelihood of acute readmission to the hospital within 6 months of hospital discharge. Currently the developed model able to correctly identify 80% of people with a who require acute readmission within 6 months of hospital discharge, surpassing an existing expert “rule-based” models.Project 2 is related to the modelling of future depression risk among newly diagnosed T1 diabetic children. This is done by using tree-based algorithms to predict the likelihood of developing a depression within 5 years of “debut” conditional on social and demographic circumstances, as well as medical history.

Kongens Lyngby, Capital Region, Denmark

I successfully defended my PhD on the 2nd of September 2022Responsibilities and activitiesConducting a long-term project and PhD-study with the aim of providing new statistical methodologies and novel findings in the field of accident analysis and safety science. This relied on time-series modelling, latent class/finite mixture models and other econometric models. Beyond working towards publishable research the work also consisted of presenting at conferences, teaching and supervising student projects. The primary, results of the PhD are as follows.Results:Developed a hybrid recurrent network and mixture density model that estimates hourly cycling traffic in Copenhagen ~77% more accurately than the model employed by the Danish Road Directorate.Publication of two research papers in peer-reviewed articles, and further two in review.

Research Assistant in the Transport Modelling group

Remote (Rome)

Responsibilities and activitiesAiding in the transition of existing econometric models used by the group to the new statistical working system and translating current Stata code to R-language. Developing web-scrapers to gather data concerning agriculture credit from national banksProposing models to improve missing data imputation in producer price data, based on machine learning gaussian processes.Results:In collaboration with the team around 35 separate web-scrapers were designed to scrape data and automatically download files containing data from various countries national bank and statistics websites.Gaussian Process regression and Matrix Factorisation were presented as viable methods to be used by the Economic Statistics Group, and some models were transitioned to the statistical working system and installed as plugins.

Köln-Området, Tyskland

Responsibilities and activitiesCollaborating with the Data Science team on providing in-house data consulting services at Real Digital, providing A.I. based and software solutions for automating decision-making processes in the other teams. Developing a data science platform for the quick launch of micro services in an easy-to-use project pipeline. Results:Creating a large model-suite to assist the customer service team in choosing templates to reply to customer complaint. Relying on various clustering and classification methodsCreation of a pipeline for micro service launches in the future, decreasing the time from development to going live with services.

Research in Big Data analysis tools (Hadoop, Spark, Kafka....), assessing strength regarding possible uses for the company. Furthermore doing automation of processes and working with realtime visualisation of data streams and other work with Advanced Analytics using Python and R

Valby

Energi- & Vandværkstedet er en innovativ og eksperimenterende miljøskole, der tilbyder undervisningsforløb for alle klassetrin inden for temaerne energi- og vandforsyning i fortid, nutid og fremtid, altid set i et bæredygtighedsperspektiv. Energi- & Vandværkstedet drives i et samarbejde mellem Miljøtjenesten i Københavns Kommune og HOFOR.

Master Of Science (Msc), Statistik

Became MSc in Statistics on the 31st of September 2017

Bachelor Of Science (B.Sc.), Fysik, Bsc

Fysik, Physics, Bsc

Activities and Societies: Udveksling gennem UCEAPUdvekslingsår