The client requested the advanced statistical analysis of their portfolio, to find out patterns, similarities and insights which could help them decrease attritional loss ratio.
Our team analyzed Company’s portfolio in detail, retrieved data insights and created a picture of the current situation. Besides that, we detected anomalies using deep autoencoder neural network and provided a distinct analysis of outliers. We have based our recommendations on how to reduce the attritional loss ratio on the results of tree ensemble models and applied loss-cost modelling approach. In addition, the portfolio was divided into clusters, so the client could realize all the patterns, similarities inside, and adjust baseline strategy with new knowledge. Besides that, the PCA analysis was also performed.
The deliverables of the project were:
A well-designed presentation that summarizes the results and analysis
Recommendations, supported by statistical significance tests, sensitivity analysis including different ways of treating large losses.
All code used to develop the analysis and all explanatory visualizations/exhibits.
As a result, the client got clusters-insights and knowledge on how to recognise weak profiles and create the opportunity to decrease the attritional loss ratio by 3%.