Advantage
Reliable forecast of key product figures
Data
Purely descriptive reporting
Method
ARIMA, Exponential Smoothing (ETS), Bayesian Structural Time Series (BSTS)
Challenge
The customer is one of Germany's largest insurance groups. The client's reinsurance department has created a Shiny app for the user-friendly preparation of key product figures such as expenses or loss ratios. The reporting generated with the app enables employees to regularly view the current status of the individual product categories. In order to be able to use the reporting even more as a basis for planning, the insurer was looking for a way to expand the existing application to include forecasting options.
Goal
The customer's aim is to implement reliable forecasting based on time series, i.e. the key product figures displayed in the app. It should be possible to output the information as graphics and tables.
Solution
Due to the broad product portfolio, the time series are very heterogeneous. Seasonalities, trends or the basic suitability for predictions: The central task for the data science specialists at eoda was therefore to evaluate the suitable analysis methods with regard to the respective time series.
In addition to the standard models ARIMA and Exponential Smoothing (ETS), eoda also examined the applicability of Bayesian Structural Time Series (BSTS).
The basic procedure for assessing the quality of a model is to divide the available data into a training and a test data set. The model is then trained with the training time series in order to be able to make predictions for the test period using this model. The comparison of the prediction with the actual course of the time series - the test data - provides information about the quality of the model.
The result is a function that recognizes the individual characteristics of the time series and uses a matrix to apply the best analysis model in the correct configuration for forecasting.
Result
With eoda's support, the insurer can expand its previously purely descriptive reporting to include reliable forecasting of product key figures. In addition to a commented analysis script, the customer also receives a detailed project report from eoda, which presents the chosen approach and the methods used. This means that the forecasting model can not only be seamlessly implemented in the existing Shiny app, but is also easy to understand for the employees responsible.
The many years of project experience and comprehensive methodological expertise of eoda's analysis specialists guaranteed fast and targeted project implementation.
Get started now:
We look forward to exchanging ideas with you.
Your expert on Data-Science-Projects:
Lutz Mastmayer
projects@eoda.de
Tel. +49 561 87948-370
