Browsing by Author "Šoltés, Erik"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
- ItemMultidimensional credibility model and its application(Technická Univerzita v Liberci, ) Pacáková, Viera; Šoltés, Erik; Linda, Bohdan; Ekonomická fakultaSolvency II project places emphasis on the modelling and management of risks of the insurance companies. This requires further improvement in actuarial methods and their application in insurance practice. Improving the quality of premium calculation methods is an effective factor in reducing the insurance technical risk of an insurer. Presentation the methods of premium calculation and its permanent updating is the aim of this article. Credibility theory is an experience rating technique to determine premiums, claim frequencies or claim sizes. Credibility models are based on the realistic concept of a heterogeneous insurance portfolio. Therefore, two sources of information are used in the calculation of the credibility estimators for the individual risk: typically little knowledge about the individual risk and quite extensive statistical information about entire portfolio. The most important model in the credibility theory is Bühlmann-Straub model. This model has a wide range of possibilities to be used in praxis mainly in general insurance. Besides that this model is a basis for other more specific models such as hierarchical, multidimensional or regression credibility models. In this article we deal with generalisation of one-dimensional Bühlmann-Straub credibility model to the multidimensional credibility model. We mainly focus on estimation of so-called structural parameters and usage of SAS Enterprise Guide application when estimating. The multidimensional Bühlmann-Straub credibility model is applied based the real data in motor vehicle third party liability insurance.
- ItemTargeting of Online Advertising Using Logistic Regression(Technická Univerzita v Liberci, ) Šoltés, Erik; Táborecká-Petrovičová, Janka; Šipoldová, Romana; Ekonomická fakultaRecently, the internet became the dominant medium in marketing and comparing the development of expenditures into advertising indicates the dominance of online advertising will be inevitably stronger. Internet advertising compared to traditional media advertising has plenty of advantages hence online marketing exhibits a huge expansion in recent era. To fully utilize the potential of online marketing, it is necessary to effectively target activities of relevant internet users with the real presumption they will purchase promoted products or services. The paper is focused on demographic targeting by the mean of logistic regression models. Explanatory variables in presented application are arising from affinities of internet webpages visited by particular users and areas of their interests that are identified from their online behaviour. Our paper provides binomial logistic mode whose role is to predict the gender of internet user and multinomial logistic model constructed for the estimation of age category the user may be assigned to. The only variables exploited in the model by the mean of stepwise regression are variables with significant influence. The impact of particular factors is quantified via odds ratios that are used for the identification of areas of interests typical for women, men and for considered age categories. The paper demonstrates how it is possible to utilise estimated logistic models for the estimation of probabilities that the internet user is from a target group – in our case, women aged 25–44 years old. Prediction quality of models is assessed by the set of classification measures arising from confusion matrix that is generally acceptable in machine learning. Presented analyses are conducted in statistical software SAS Enterprise Guide on data provided from the real advertising campaign. More than 160,000 statistical units enabled the confirm results gained on training dataset of a relatively huge validation dataset.