Individual Loss Reserving Using a Gradient Boosting-Based Approach

Duval, Francis; Pigeon, Mathieu

doi:10.3390/risks7030079

Cited by 23 publications

(5 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The author concluded that the gradient boosting algorithm provided a better performance than DT and GP and also exhibited that it can effortlessly classify and predict the number of dry, normal, and wet events in both case studies. Pigeon and Duval [21] also used the gradient boosting algorithm along with the generalized linear models in predicting the total paid amount of each claim in insurance. It is also claimed that GBA is an efficient method on computing the predicted values such as total amount paid for each claim and payment schedule.…”

Section: Related Workmentioning

confidence: 99%

Detection of Earthquake Damages from Satellite Images using Gradient Boosting Algorithm with Decision Trees as Base Estimator

Acula¹

2022

actamanil

View full text Add to dashboard Cite

The Philippines experiences an average of twenty (20) earthquakes every day and 100 to 150 felts every year. With the aim to help the government in mitigation of the potential impact of the earthquake in the country, this research explored the detection of earthquake damages from the satellite images. Since Gradient Boosting Algorithm is considered one of the efficient and powerful predictive models, this method was employed in this research to classify if the satellite images before and after the earthquake brought damages in the infrastructure of affected areas. The satellite images used in the research were downloaded from Landsat 8 via Google Earth pro, which focused on a Magnitude 7.6 earthquake in Eastern Samar on August 31, 2012. The research used image augmentation such as rotation, shearing and flipping before the image extraction. The extracted features were used as classifiers of the models and were separated into 80:20, 70:30 and 60:20 ratios for training and testing sets respectively. The detection of damages was evaluated five (5) times using different n-estimators or numbers of trees {10, 20, 30, 40, 50}. The experiment concluded that Gradient Boosting Algorithm is an efficient model for classification and detection of earthquake damages using satellite image data, with an optimal detection accuracy of 85.71% and 100% without and with image augmentation.

show abstract

Section: Related Workmentioning

confidence: 99%

Detection of Earthquake Damages from Satellite Images using Gradient Boosting Algorithm with Decision Trees as Base Estimator

Acula¹

2022

actamanil

View full text Add to dashboard Cite

show abstract

“…They proposed a granular model for the heterogeneity in the observation delay based on the occurrence day of the event and on calendar day effects in the observation process, such as weekday and holiday effects. Duval and Pigeon (2019) modelled loss reserving for non-life insurance by combining traditional approaches and gradient boosting algorithms in an individual framework. Their gradient boosting models allowed insurers to compute a prediction for the total paid amount of each claim.…”

Section: Introductionmentioning

confidence: 99%

Prediction of IBNR and RBNS Liabilities using Estimated Delay Probabilities by a Zero-Inflated Gamma Mixture Model

Atatalab,

Payandeh

2024

Preprint

View full text Add to dashboard Cite

This article considers the problem of predicting the amount/number of claims that have been incurred but not reported, say IBNR, and reported but not settled, say RBNS. Using the delay probabilities, calculated by the ZIGM distribution, it predicts both IBNR’s and RBNS’ loss reserves and the corresponding mean square error. The practical application of our findings are applied against a real data.

show abstract

“…Wüthrich 14 used regression trees to predict the number of payments. Tree‐based techniques like ExtraTrees and XGBoost have been applied to predict outstanding individual losses, see, for example, References 15 and 16.…”

Section: Introductionmentioning

confidence: 99%

Micro‐level reserving for general insurance claims using a long short‐term memory network

Chaoubi

Besse²,

Cossette

et al. 2023

Appl Stoch Models Bus & Ind

View full text Add to dashboard Cite

Detailed information about individual claims are completely ignored when insurance claims data are aggregated and structured in development triangles for loss reserving. In the hope of extracting predictive power from the individual claims characteristics, researchers have recently proposed to use micro‐level loss reserving approaches. We introduce a discrete‐time individual reserving framework incorporating granular information in a deep learning approach named Long Short‐Term Memory (LSTM) neural network. At each time period, the network has two tasks: first, classifying whether there is a payment or a recovery, and second, predicting the corresponding non‐zero amount, if any. Based on a generalized Pareto model for excess payments over a threshold, we adjust the LSTM reserve prediction to account for extreme payments. We illustrate the estimation procedure on a simulated and a real general insurance dataset. We compare our approach with the chain‐ladder aggregate method using the predictive outstanding loss estimates and their actual values.

show abstract

Individual Loss Reserving Using a Gradient Boosting-Based Approach

Cited by 23 publications

References 25 publications

Detection of Earthquake Damages from Satellite Images using Gradient Boosting Algorithm with Decision Trees as Base Estimator

Detection of Earthquake Damages from Satellite Images using Gradient Boosting Algorithm with Decision Trees as Base Estimator

Prediction of IBNR and RBNS Liabilities using Estimated Delay Probabilities by a Zero-Inflated Gamma Mixture Model

Micro‐level reserving for general insurance claims using a long short‐term memory network

Contact Info

Product

Resources

About