A survey of personalized treatment models for pricing strategies in insurance

Guelman, Leo; Guillén, Montserrat; Pérez-Marín, Ana M.

doi:10.1016/j.insmatheco.2014.06.009

Cited by 28 publications

(25 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Ada: Square loss with n estimators=100. [5,10,20,40]. Parameter selection is conducted by validation because of the time constraint.…”

Section: Appendixmentioning

confidence: 99%

“…A new subject is simply assigned to the empirically best treatment as measured on the K training data that are closest to it. Several tree-based algorithms have been proposed for uplift modeling, each with a different splitting criterion [8] [1] [9] [4] [10]. In [8], the authors modify the standard decision tree construction procedure [11] by forcing a split on the treatment at each leaf node.…”

Section: Introductionmentioning

confidence: 99%

“…In [4], a linear model is fitted to each candidate split and the significance of the interaction term is used as the measure of the split quality. In [10], the variable that has the smallest p-value in the hypothesis test on the interaction between the response and itself is selected as the splitting variable. Then the splitting point is chosen to maximize the interaction effect.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Uplift Modeling with Multiple Treatments and General Response Types

Zhao¹,

Xiao²,

Simchi‐Levi³

2017

Proceedings of the 2017 SIAM International Conference on Data Mining

View full text Add to dashboard Cite

Randomized experiments have been used to assist decisionmaking in many areas. They help people select the optimal treatment for the test population with certain statistical guarantee. However, subjects can show significant heterogeneity in response to treatments. The problem of customizing treatment assignment based on subject characteristics is known as uplift modeling, differential response analysis, or personalized treatment learning in literature. A key feature for uplift modeling is that the data is unlabeled. It is impossible to know whether the chosen treatment is optimal for an individual subject because response under alternative treatments is unobserved. This presents a challenge to both the training and the evaluation of uplift models. In this paper we describe how to obtain an unbiased estimate of the key performance metric of an uplift model, the expected response. We present a new uplift algorithm which creates a forest of randomized trees. The trees are built with a splitting criterion designed to directly optimize their uplift performance based on the proposed evaluation method. Both the evaluation method and the algorithm apply to arbitrary number of treatments and general response types. Experimental results on synthetic data and industry-provided data show that our algorithm leads to significant performance improvement over other applicable methods.

show abstract

“…Ada: Square loss with n estimators=100. [5,10,20,40]. Parameter selection is conducted by validation because of the time constraint.…”

Section: Appendixmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Uplift Modeling with Multiple Treatments and General Response Types

Zhao¹,

Xiao²,

Simchi‐Levi³

2017

Proceedings of the 2017 SIAM International Conference on Data Mining

View full text Add to dashboard Cite

show abstract

“…The performance of UCTS is compared with those of 6 other methods which are the separate model approach with Random Forest (SMA-RF), Support Vector Machine (SMA-SVM), Adaboost (SMA-Ada), K-Nearest Neighbors (SMA-KNN), as well as the uplift Random Forest method implemented in [11], and CTS. The data is randomly split into the training set (225,000 samples per treatment) and the test set (75,000 samples per treatment).…”

Section: Priority Boarding Datamentioning

confidence: 99%

A Practically Competitive and Provably Consistent Algorithm for Uplift Modeling

Zhao

Xiao

Simchi‐Levi

2017

2017 IEEE International Conference on Data Mining (ICDM)

View full text Add to dashboard Cite

Randomized experiments have been critical tools of decision making for decades. However, subjects can show significant heterogeneity in response to treatments in many important applications. Therefore it is not enough to simply know which treatment is optimal for the entire population. What we need is a model that correctly customize treatment assignment base on subject characteristics. The problem of constructing such models from randomized experiments data is known as Uplift Modeling in the literature. Many algorithms have been proposed for uplift modeling and some have generated promising results on various data sets. Yet little is known about the theoretical properties of these algorithms. In this paper, we propose a new tree-based ensemble algorithm for uplift modeling. Experiments show that our algorithm can achieve competitive results on both synthetic and industry-provided data. In addition, by properly tuning the "node size" parameter, our algorithm is proved to be consistent under mild regularity conditions. This is the first consistent algorithm for uplift modeling that we are aware of.

show abstract

“…They show how models of personalized treatment learning can be used to select the policy holders that should be targeted in marketing strategies of a company [28].…”

Section: Related Workmentioning

confidence: 99%

Predicting Behaviors Of Insurance Costumers By Using The Genetic Algorithm

Bagheri

Ziaratban

Tarokh

2015

J. Math. Computer Sci.

View full text Add to dashboard Cite

Improving customers' satisfaction, retaining loyal customers, and acquiring new customers are the most important goals of companies. Because of existing competition conditions among insurance industries, especially in optional insurances such as vehicle body insurance, implementing policies which have benefits for both customers and insurance companies can increase the corporation profitability and customers' satisfaction. In this paper, general policies of vehicle body insurance in Iran are investigated. In this way, a non-linear function (Cj) is proposed to separate profitable and harmful customers. This function is based on some insurance features of customers such as age of a vehicle, vehicle price, vehicle type, and number of years without accident. Customers are categorized based on the calculated values of Cj in which all parameters are optimally set. The parameters have to be set in the training phase to calculate reliable Cj values for customers. Parameters of Cj are set so that in a subset of the customers of the training set for which the value of Cj is greater than a threshold (Th), the number of profitable customers and disadvantageous customers should be maximized and minimized, respectively. A customer with positive M (monetory) value is a profitable customer. To optimize this multivariable problem with large number of parameters, the genetic algorithm is used. After optimizing the parameters in the training phase by using GA, the behaviors of the customers in the test set are predicted. In the prediction phase, it was estimated that a costumer is profitable or not, even he had accidents and received recompenses from the company. If a customer is predicted as profitable, the company allows him to enjoy some discounts. In the current Iranian insurance policies, all discounts are canceled for next insurance year if a customer has accidents and gives recompense from the company. By implementing the proposed policy, the customers' satisfactions are increased and also more new customers are acquired from other insurance companies.

show abstract

A survey of personalized treatment models for pricing strategies in insurance

Cited by 28 publications

References 40 publications

Uplift Modeling with Multiple Treatments and General Response Types

Uplift Modeling with Multiple Treatments and General Response Types

A Practically Competitive and Provably Consistent Algorithm for Uplift Modeling

Predicting Behaviors Of Insurance Costumers By Using The Genetic Algorithm

Contact Info

Product

Resources

About