XGBoost is recognized as an algorithm with exceptional predictive capacity. Models for a binary response indicating the existence of accident claims versus no claims can be used to identify the determinants of traffic accidents. This study compared the relative performances of logistic regression and XGBoost approaches for predicting the existence of accident claims using telematics data. The dataset contained information from an insurance company about the individuals' driving patterns-including total annual distance driven and percentage of total distance driven in urban areas. Our findings showed that logistic regression is a suitable model given its interpretability and good predictive capacity. XGBoost requires numerous model-tuning procedures to match the predictive performance of the logistic regression model and greater effort as regards to interpretation.Risks 2019, 7, 70 2 of 16 among many others. Here, pay-as-you-drive (PAYD) insurance schemes represent an alternative method for pricing premiums based on personal mileage travelled and driving behaviors. Guillen et al. (2019), Verbelen et al. (2018, and Pérez-Marín and Guillén (2019) showed the potential benefits of analyzing telematics information when calculating motor insurance premiums. Gao and Wüthrich (2019) analyzed high-frequency GPS location data (second per second) of individual car drivers and trips. Gao and Wüthrich (2018) and investigated the predictive power of covariates extracted from telematics car driving data using the speed-acceleration heatmaps proposed by Wüthrich (2017). Further, Hultkrantz et al. (2012) highlighted the importance of PAYD insurance plans insofar as they allow insurance companies to personalize premium calculation and, so, charge fairer rates.The rest of this paper is organized as follows. First, the notation is introduced and the logistic regression and XGBoost methods are outlined. Second, our dataset is described and some descriptive statistics are provided. Third, the results of our comparisons in both a training and a testing sample are reported. Finally, following the conclusion, some practical suggestions are offered about the feasibility of applying new machine learning methods to the field of insurance.
Transmucosal delivery is commonly used to prevent or treat local diseases. Pranoprofen is an anti-inflammatory drug prescribed in postoperative cataract surgery, intraocular lens implantation, chorioretinopathy, uveitis, age-related macular degeneration or cystoid macular edema. Pranoprofen can also be used for acute and chronic management of osteoarthritis and rheumatoid arthritis. Quality by Design (QbD) provides a systematic approach to drug development and maps the influence of the formulation components. The aim of this work was to develop and optimize a nanostructured lipid carrier by means of the QbD and factorial design suitable for the topical management of inflammatory processes on mucosal tissues. To this end, the nanoparticles loading pranoprofen were prepared by a high-pressure homogenization technique with Tween 80 as stabilizer and Lanette® 18 as the solid lipid. From, the factorial design results, the PF-NLCs-N6 formulation showed the most suitable characteristics, which was selected for further studies. The permeability capacity of pranoprofen loaded in the lipid-based nanoparticles was evaluated by ex vivo transmucosal permeation tests, including buccal, sublingual, nasal, vaginal, corneal and scleral mucosae. The results revealed high permeation and retention of pranoprofen in all the tissues tested. According to the predicted plasma concentration at the steady-state, no systemic effects would be expected, any neither were any signs of ocular irritancy observed from the optimized formulation when tested by the HET-CAM technique. Hence, the optimized formulation (PF-NLCs-N6) may offer a safe and attractive nanotechnological tool in topical treatment of local inflammation on mucosal diseases.
Pranoprofen (PRA)-loaded nanostructured lipid carriers (NLC) have been dispersed into blank gels composed of 1% of Carbomer 940 (PRA-NLC-Car) and 3% of Sepigel® 305 (PRA-NLC-Sep) as a novel strategy to refine the biopharmaceutical profile of PRA, for dermal administration in the treatment of skin inflammation that may be caused by possible skin abrasion. This stratagem intends to improve the joining of PRA with the skin, improving its retention and anti-inflammatory effect. Gels were evaluated for various parameters such as pH, morphology, rheology, and swelling. In vitro drug release research and ex vivo permeation through the skin were carried out on Franz diffusion cells. Additionally, in vivo assays were carried out to evaluate the anti-inflammatory effect, and tolerance studies were performed in humans by evaluating the biomechanical properties. Results showed a rheological profile common of semi-solid pharmaceutical forms for dermal application, with sustained release up to 24 h. In vivo studies using PRA-NLC-Car and PRA-NLC-Sep in Mus musculus mice and hairless rats histologically demonstrated their efficacy in an inflammatory animal model study. No signs of skin irritation or modifications of the skin’s biophysical properties were identified and the gels were well tolerated. The results obtained from this investigation concluded that the developed semi-solid formulations represent a fitting drug delivery carrier for PRA’s transdermal delivery, enhancing its dermal retention and suggesting that they can be utilized as an interesting and effective topical treatment for local skin inflammation caused by a possible abrasion.
Historical records of insured policy holders constitute an ideal environment for the development of machine learning algorithms. These procedures are implemented on databases in order to extract knowledge. Here we explore different approaches to the prediction of claims and premiums in the automotive industry, comparing their implementation in a real sample, randomly divided into training and testing. We propose measures to help in the evaluation of the methods and their practical implication for the prediction of rare events and premium calculation. The main conclusion is that the dispersion of prices, and specifically the difference between the maximum pure and minimum premium, can become very different according to the predictive method used. Keywords: data science, artificial intelligence, nonlife insurance, premiums, claims Resumen La información histórica de los asegurados constituye un entorno idóneo para el desarrollo de los algoritmos machine learning, cuya finalidad es extraer conocimiento a partir de bases de datos. En este artículo exploramos diversas aproximaciones a la predicción de la siniestralidad y las primas del ramo del automóvil, comparando su implementación en una muestra real, dividida aleatoriamente en muestras de entrenamiento y test. Proponemos medidas para ayudar en la valoración de los métodos y su implicación práctica para la predicción de eventos pocos frecuentes y el cálculo de primas. La principal conclusión es que la dispersión de precios, y concretamente la diferencia entre la prima pura máxima y mínima, puede llegar a ser muy diferente según el método predictivo utilizado. Palabras clave: ciencia de datos, inteligencia artificial, seguros generales, primas, siniestros
A boosting-based machine learning algorithm is presented to model a binary response with large imbalance, i.e., a rare event. The new method (i) reduces the prediction error of the rare class, and (ii) approximates an econometric model that allows interpretability. RiskLogitboost regression includes a weighting mechanism that oversamples or undersamples observations according to their misclassification likelihood and a generalized least squares bias correction strategy to reduce the prediction error. An illustration using a real French third-party liability motor insurance data set is presented. The results show that RiskLogitboost regression improves the rate of detection of rare events compared to some boosting-based and tree-based algorithms and some existing methods designed to treat imbalanced responses.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.