The feasibility of adding machine learning in the manual medical underwriting process at If Insurance
Saarinen, Meri (2021-06-17)
The feasibility of adding machine learning in the manual medical underwriting process at If Insurance
Saarinen, Meri
(17.06.2021)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
suljettu
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe2021061838889
https://urn.fi/URN:NBN:fi-fe2021061838889
Tiivistelmä
This thesis aims to find if it is feasible to implement machine learning to manual underwriting process at If Insurance. The research uses both qualitative and quantitative research methods. The qualitative research methods are interviews that aim to answer if machine learning can benefit the risk assessors. The quantitative methods are research planning, preprocessing the data, training machine learning algorithms and analysing the results. The data used in the algorithms is pseudonymised If Insurance electronic health declaration data from Norway. Approximately 8000 data points are used in the thesis. The data contains both structured and text data. The machine learning algorithms are used to discover the potential of using the data to reach automation levels good enough to be used in any actual environment. Ultimately it seems that adding machine learning to manual handling is feasible. The feasibility is based on the high F1-scores of the machine learning algorithms and the risk assessor interviews. The interviews concluded that the risk assessors felt a benefit in adding machine learning to manual handling. To further advance the research, more data should be acquired. Acquiring the missing data will be possible in the future when the data that is not currently stored is accessible.