Service Analytics
Permanent URI for this collection
Browse
Recent Submissions
1 - 4 of 4
-
ItemUsage Space Sampling for Fringe Customer Identification( 2021-01-05)With large numbers of available customers, it is often essential to select representative samples for reasons of computational cost reduction and upstream advanced data analytics. However, for many analytical procedures, the usage behavior observed from a smaller sample of customers must indicate well the fringe of usage and its relation to extreme product loads. Due to the high complexity of technical or service systems, it remains challenging to minimize the number of samples while sufficiently capturing the fringe customers. With the availability of data related to usage behavior, we consider a sampling method to address this problem by analyzing the customer usage space before sampling, then separately sampling fringe and core customers, and weighting the samples afterwards. Experimental results show that the method can identify fringe customers with significantly fewer, yet reproducible samples, while maintaining the distribution representativeness of customer population to a large extend.
-
ItemUpgrading Products based on Existing Dominant Competitors( 2021-01-05)In the Industry 4.0 era, manufacturers compete to produce better products that are expected to satisfy a larger number of customers. We propose a recommendation system for upgrading products considering user preferences. This approach is based on the dominating regions of dominant competitors. The dominating region represents the estimation of the number of potential customers. However, examining overlapped dominating regions for a high dimensional space is NP-hard. We propose a novel method named TDRDFS which constructs a Dominant Graph of Intersection skyline points (DGI) for modeling the dominating regions. Our experiments show that TDRDFS significantly reduces computation. Based on our approach, product vendors are able to determine the strategy of upgrading products easily.
-
ItemTowards a Machine Learning-based Decision Support System for Dispatching Helicopters in New Zealand( 2021-01-05)Helicopters play an important role in emergency medical service systems worldwide. In sparsely populated countries like New Zealand with long distances between hospitals, helicopters are often the best way to help critically injured patients. As helicopters are extremely costly, they should only be dispatched when really necessary. In this paper, we use data from the South Island of New Zealand to test several Machine Learning approaches and show that they can be used to support dispatchers by identifying emergencies likely to require a helicopter response. We follow a non-static dataset, as the information is successively available during an emergency, and demonstrate that even a limited approach, based only on geographic incident information, can yield an Average Precision of 94% for highlighting critical emergencies. In the latter parts of this paper, we investigate different compositions of training data to assess the impact of a potential concept drift.
-
ItemIntroduction to the Minitrack on Service Analytics( 2021-01-05)