Improved disability insurance model with collective health claims.
problem Enhance disability insurance model with collective health claims.
method Expand classic semi-Markov model with collective health claims, solve many-body problem using mean-field approach.
result Mean-field approach simplifies complex model into a transparent pricing method.
Two machine learning models detect anomalies in ER claims, saving up to 40% in improper payments.
problem Improper health insurance payments from fraud and upcoding.
method Two machine learning models: an upcoding model based on severity code distributions and a random forest model for claim sorting.
result Random forest model saved 12% to 40% in improper payments compared to a baseline approach.
Deep neural network predicts health costs better than traditional models.
problem Accurate prediction of healthcare costs for optimal cost management.
method Developed a deep neural network to predict future health care costs from health insurance claims records.
result Deep neural network outperformed ridge regression and Morbi-RSA models in cost prediction.
Deep learning text embeddings improve fraud detection in healthcare insurance.
problem Improving fraud detection in healthcare insurance claims.
method Proposed deep learning architectures for text embeddings.
result Our approach outperforms other methods in detecting fraudulent claims.
Insurance companies must manage millions of claims per year. While most of these claims are non-fraudulent, fraud detection is core for insurance companies. The ultimate goal is a predictive model to single out the fraudulent claims and pay out the non-fraudulent ones immediately. Modern machine learning methods are we…
Enhanced Tweedie model for insurance claims using CatBoost.
problem Accurately modeling aggregate claims with zero-inflated data.
method Refined Tweedie model with boosting methods in CatBoost.
result Marked improvement in model performance for insurance analytics.
Deep Claim predicts payer responses from claims data using deep learning.
problem Predicting payer responses from claims data to improve healthcare performance.
method Learning complex dependencies in claim inputs to create a compact representation, then using deep learning to predict responses.
result Deep Claim improves claim denial prediction by 22.21%.
Study predicts high-cost patients using insurance claims data.
problem Accurately identifying high-cost patients to manage costs.
method Applied machine learning to health insurance claims data.
result Developed a high-performance algorithm with 91.2% AUC.
Study assesses health plan risk measures for Solvency Capital Requirement.
problem Assessing risk measures for health plans to meet Solvency Capital Requirement.
method Three-part regression model with three GLMs for claim counts, episode allocation, and severity.
result Reduction in regression models compared to traditional methods.
Model predicts individual insurance claim reserves using activation patterns.
problem Accurately predicting individual claim reserves in insurance contracts.
method Multinomial logistic regression to model claim activation and development.
result The model generates accurate predictions of total and per coverage reserves.
Study tackles imbalanced data in car insurance claims prediction.
problem Predicting rare events (claims) in car insurance with imbalanced data.
method Various machine learning techniques (logistic-regression, decision tree, random forest, xgBoost, feed-forward network) applied to imbalanced dataset.
result Comparison of machine learning algorithms' performance in claim occurrence prediction.
Study uses SVM to predict weather-induced home insurance claims and losses.
problem Assessing future weather-induced home insurance claims and losses for disaster preparedness.
method Support Vector Machine (SVM) regression for forecasting future claim dynamics.
result Illustrates SVM approach in forecasting weather-induced home insurance claims in a Canadian city.
New model bridges pricing and reserving for insurance claims.
problem Incomplete claim data due to reporting and settlement delays.
method Develops an occurrence and development model to estimate both claims and premiums.
result Effective resolution of pricing and reserving inconsistencies.
Model detects insurance fraud using social network analysis.
problem Fraudulent insurance claims by exaggeration or intentional damage.
method Network construction linking claims and parties, BiRank algorithm for fraud score computation, feature extraction from network and claims, supervised model building.
result Network features improve fraud detection performance.
Study compares ruin probabilities under independence vs. dependence assumptions.
problem Underestimation of ruin probability when claims are dependent.
method Copulas for claim dependence analysis, sensitivity analysis.
result Dependent claims lead to underestimation of ruin probability.
Optimizes insurance processing capacity to minimize costs.
problem Processing delays and backlogs in insurance claims.
method Optimal capacity selection to minimize delay-adjusted and fixed costs.
result Minimizes claims costs by balancing processing capacity and delays.
Bayesian CART models improve insurance claims frequency prediction and interpretation.
problem Improving accuracy and interpretability in insurance pricing models.
method Introducing Bayesian CART models for claims frequency, implementing MCMC algorithm for posterior tree exploration, and using DIC for model selection.
result Bayesian CART models can better classify policy-holders into risk groups.
Machine learning models outperform traditional actuarial methods in predicting health insurance costs.
problem Improving accuracy in health insurance pricing to identify concession opportunities.
method Developed and evaluated two machine learning models at the patient and employer-group levels.
result Machine learning models outperformed traditional actuarial models by 20% in predicting costs.
This study compares the largest claims from two insurance portfolios using stochastic orderings.
problem Comparing the largest claims from two heterogeneous insurance portfolios.
method Used various stochastic orderings and established sufficient conditions associated with model parameters.
result Established sufficient conditions for comparing the largest claims from two insurance portfolios.
The article proposes a method to make valid insurance claim predictions without relying on specific models.
problem Prediction of insurance claims using statistical models can be unreliable due to model misspecification, selection effects, and lack of finite-sample validity.
method The article employs conformal prediction, a machine learning strategy that is model-free and tuning-parameter-free, ensuring finite-sample validity.
result The proposed method guarantees valid predictions at a pre-assigned coverage probability level and performs well in insurance applications, including meeting Solvency II requirements.
This paper suggests claim history will be deprecated in future auto insurance rates.
problem The role of historical claim records in auto insurance rates.
method Proposes a new risk variable elimination method and real-time road risk model design.
result Claim history will be considered a 'noise' factor and deprecated in Pay-How-You-Drive models.
It is illustrated a methodology to compute the pure premium for the automobile insurance (claim frequency and severity) using generalized linear models. It is obtained the pure premium for the partial damage loss cover (PPD) using a set of automobile insurance policies with an exposition of a year. It is found that the…
Detects organized fraudsters in insurance claims with high precision.
problem Fraudulent insurance claims lead to heavy financial losses.
method Developed a novel data-driven procedure using graph learning algorithms.
result Achieves more than 80% precision in fraud detection.
The distribution of health care payments to insurance plans has substantial consequences for social policy. Risk adjustment formulas predict spending in health insurance markets in order to provide fair benefits and health care coverage for all enrollees, regardless of their health status. Unfortunately, current risk a…
Study classifies liability insurance policies using machine learning.
problem Classifying liability insurance policies with or without claims.
method Used machine learning models like nearest neighbour and logistic regression on Actuarial Challenge dataset.
result Models accurately classified policies into claims and non-claims groups.
A new method for modeling insurance claim frequencies using random proportions.
problem Inaccurate fitting of classical distributions to insurance claim frequency data.
method Modeling claim frequencies using random proportions of insurance contracts and applying goodness-of-fit tests.
result A new statistical approach for better modeling insurance claim frequencies.
The study analyzes how bonus-malus systems and delayed claims settlement affect insurance companies' financial stability.
problem Analyzing the impact of bonus-malus systems and delayed claims settlement on insurance companies' financial stability.
method Examined a discrete-time risk model with time-varying premiums, evaluating two types of claims and settlement delays.
result Delayed settlement of by-claims leads to lower ruin probabilities under specific assumptions.
The paper analyzes systemic risk in an insurance model with multiple business lines and heterogeneous claims.
problem Analyzing systemic risk in a multi-dimensional insurance model with heterogeneous claims.
method A multi-dimensional Lévy process-based renewal risk model with pairwise asymptotic independence (PAI).
result Asymptotic formulas for tail probabilities and systemic risk measures are derived.
The paper proposes an original methodology for constructing quantitative statistical models based on multidimensional distribution functions constructed on the basis of the insurance companies' data on inshurance policies (including policies with deductible) and claims incurred. Real data of some Russian insurance comp…
New models for insurance claims accounting for delays.
problem Delayed claims processing affects real-time policy development.
method Introducing transaction time models to describe multi-state life insurance.
result Explicit expressions for transaction time reserves derived.
EBM improves car insurance claim severity and frequency prediction while maintaining interpretability.
problem Balancing predictive accuracy and interpretability in insurance claim modeling.
method Combines GAM and cyclic gradient boosting, providing interpretable predictions.
result EBM outperforms benchmark models in claim severity and frequency prediction.
New model for disability insurance reserving handles delays in claim information.
problem Disability insurance claims are affected by long delays and adjudication processes.
method Proposes a new individual reserving model for real-time claim evolution.
result Shows that new reserves can be calculated as modifications of classic reserves.
Reinforcement learning improves insurance claims reserving by learning from all claim trajectories.
problem Traditional reserving models learn only from settled claims, missing valuable data from ongoing claims.
method Formulated as a Markov decision process, uses reinforcement learning to update OCL estimates sequentially.
result Soft Actor-Critic implementation achieves competitive claim-level accuracy and strong aggregate performance.
New methods for quantifying insurance claim cost uncertainty using LightGBM and GLMs.
problem Quantifying prediction uncertainty in insurance claim costs.
method Proposed non-conformity measures for GLMs and GBMs with Tweedie loss.
result Locally weighted Pearson residuals outperform other methods in maintaining nominal coverage with smallest average width.
Enhances insurance loss models using InsurTech data and machine learning.
problem Traditional insurance loss models lack predictive accuracy due to limited data sources.
method Combining proprietary claims data with InsurTech data and applying machine learning techniques.
result Improved predictive accuracy of the loss model through machine learning.
Paper defines AI-specific loss reconstruction problem and introduces CER framework.
problem Reconstructing AI-generated losses, especially in agentic systems.
method CER framework: C (control boundary), E (evidence reconstruction), R (insurance response).
result Defines AI-specific reconstruction problem and operationalizes it.
This note fills the gap in market-consistent valuation of lifelong health insurance products.
problem Market-consistent valuation of lifelong health insurance products is not well-addressed.
method Constructs a valuation portfolio to separate Best Estimate into policy data and financial instrument prices.
result The Best Estimate valuation is not uniquely determined by prevailing term structures and requires a stochastic model.
A new model integrates claim history and covariates for non-life insurance reserving.
problem Lack of information in traditional reserving models for non-life insurance claims.
method Hierarchical reserving model integrating claim history and covariates.
result Flexibility and robustness of the hierarchical reserving model demonstrated.
Paper proposes optimal investment and reinsurance strategies considering financial and insurance risks dependence.
problem Optimal investment and reinsurance strategies under dependent financial and insurance risks.
method Stochastic control approach to maximize expected exponential utility of terminal wealth.
result Minimal dependence between financial and insurance risks significantly impacts investment and reinsurance strategies.
Study optimal dividend strategies for insurers with natural catastrophe claims.
problem Maximizing dividends for a catastrophe insurer over its lifetime.
method Two-dimensional stochastic control problem, viscosity solutions, numerical approximation.
result Optimal dividend strategies identified for natural catastrophe insurers.
In order to submit a claim to insurance companies, a doctor needs to code a patient encounter with both the diagnosis (ICDs) and procedures performed (CPTs) in an Electronic Health Record (EHR). Identifying and applying relevant procedures code is a cumbersome and time-consuming task as a doctor has to choose from arou…
InfDetect detects e-commerce insurance fraud using graph analysis.
problem Detecting fraudulent claims in e-commerce insurance with multiple parties involved.
method Developed a large-scale fraud detection system InfDetect using graph-based approaches.
result InfDetect successfully detected thousands of fraudulent claims and saved money daily.
The paper models SaaS products as insurance, offering new pricing tools.
problem Modeling capped-usage SaaS products with insurance principles.
method Frequency-severity decomposition, premium calculation, Monte Carlo simulations.
result SaaS pricing can be analyzed using insurance actuarial methods.
Paper studies pricing and hedging of nonreplicable insurance contracts using benchmark-neutral approach.
problem Pricing and hedging of long-term insurance contracts like variable annuities.
method Benchmark-neutral pricing framework using stock growth optimal portfolio as numéraire.
result Prices can be significantly lower than risk-neutral ones, offering attractive long-term risk-management.
A new method for predicting insurance claims with statistical guarantees.
problem Creating accurate prediction intervals for insurance claims.
method Model-agnostic framework using split conformal prediction for frequency-severity modeling.
result Shows effectiveness on simulated and real datasets using various models.
It is not clear how to target patients who are most likely to benefit from digital care management programs ex-ante, a shortcoming of current risk score based approaches. This study focuses on defining impactability by identifying those patients most likely to benefit from technology enabled care management, delivered …
We study the application of dynamic pricing to insurance. We view this as an online revenue management problem where the insurance company looks to set prices to optimize the long-run revenue from selling a new insurance product. We develop two pricing models: an adaptive Generalized Linear Model (GLM) and an adaptive …
Optimizes insurance pricing to minimize ruin probability under various claim dependencies.
problem Determining optimal insurance premiums in the presence of dependencies between claim occurrences.
method Analyzes both independent and dependent claim processes, considering single and multiple risks.
result Optimal insurance premiums depend on initial reserve and claim dependencies.