Machine Learning Methods in Statistical Analysis Conducted by Statistikkonsulterna
Isabel Felix-chan
on
May 7, 2024
Statistikkonsulterna has 25 years experience of statistical analysis, using both data collected for primary analysis and register data. Methods in ML include regression and correlation methods, random forests, PCA, cluster analysis and other data driven methods. Statistikkonsulterna is currently in the process of analyzing a unique cohort of elderly in Sweden focusing on self harm. They are, in cooperation with two master students, evaluating the efficiency and precision in ML versus classical methods.
Case Study
Statistikkonsulterna conducted an internal pilot study where they predicted national migration patterns in Sweden using statistical methods and machine learning applied on publicly available data. The results were presented by European Network for Business and Industry Statistics. The results were sponsored by the Swedish innovation agency (VINNOVA). [1]
The goal of the study was to create a model for predicting migration patterns within and between municipalities in Sweden and data was retrieved from public data sources. The gravity model was applied using methods with AI/ML and compared with “traditional” regression models. The study found that while a traditional (gravity) model is more accurate than a neural network, their performance approach each other when more data is acquired.
An example of the visualization of results using an Artificial Neural Networks and a traditional Gravity model is shown below.
Current Projects
Statistikkonsulterna is cooperating with Halmstad University in a project Predicting Suicide Risk in Older Adults with decision transparency using XAI. They have applied and compared classical statistical methods and machine learning to a cohort of 1.4 million elderly people in Sweden. Data is provided from Swedish National Board of Health and Welfare. The project is expected to be reported in June 2024. [2]
References
- Pavia, Olofsson and Pettersson (2021). Predicting migration patterns in Sweden using a gravity model and neural networks, ENBIS-21 online conference.
- von Hacht and Karlsson (2024). Predicting Suicide Risk in Older Adults with decision transparency using XAI, expected in 2024.