We present the first empirical study of the impact of loss-related errors on TV viewing engagement across disparate platforms, delivery technologies and performance measures. Our dataset comprises anonymized video viewing sessions and data about quality of delivery from a content service provider with a nationwide customer base. We study buffering events on streaming apps, mild and severe packet loss errors on multicast-delivered IPTV to a Set-Top-Box (STB) and signal strength errors on Digital Terrestrial TV. Since these metrics cannot be directly compared to each other, we use engagement as our proxy measure. We first characterize the relationship between each impairment and viewing engagement, investigating confounding factors such as type of content, asset length and connection type. We conclude that the loss of engagement due to poor quality delivery is incurred immediately for on-demand content and in the long-term for live content. We rank impairments across platforms by their impact on engagement.

Abstract Large service companies run complex customer service processes to provide communication services to their customers. The flawless execution of these processes is essential because customer service is an important differentiator. They must also be able to predict if processes will complete successfully or run into exceptions in order to intervene at the right time, preempt problems and maintain customer service. Business process data are sequential in nature and can be very diverse. Thus, there is a need for an efficient sequential forecasting methodology that can cope with this diversity. This paper proposes two approaches, a sequential k nearest neighbour and an extension of Markov models both with an added component based on sequence alignment. The proposed approaches exploit temporal categorical features of the data to predict the process next steps using higher order Markov models and the process outcomes using sequence alignment technique. The diversity aspect of the data is also added by considering subsets of similar process sequences based on k nearest neighbours. We have shown, via a set of experiments, that our sequential k nearest neighbour offers better results when compared with the original ones; our extension Markov model outperforms random guess, Markov models and hidden Markov models.

The efficient management of networks and the provisioning of services with desired QoS guarantees is a challenge which needs to be addressed through autonomous mechanisms which are intelligent, lightweight and scalable. Recent focus on applying Machine Learning approaches to model the network and service behavioural patterns have proved to be quite effective in fulfilling the objectives of autonomous management. To this end, this paper advances on the idea of implementing a distributed management solution which harnesses the predictive capability of Bayesian Networks (BN). A multi-node distributed Call Admission Control solution (termed as BNDAC) is proposed and implemented to demonstrate the modelling and prediction power of BN. A thorough evaluation of BNDAC is presented in terms of its prediction accuracy, algorithmic complexity and decision-making speed. In an online setup, performance of BNDAC is evaluated and compared with a centralised scenario, to demonstrate its superior performance for Call Blocking Probability and QoS provisioning. Simulation results based on Opnet Modeler and Hugin Researcher show the feasibility and applicability of BNDAC solution for real-time operation and management of real world networks such as the NGN.

The advent of IP-based Next Generation Network (NGN) and its guaranteed QoS promise has attracted significant attention from both service providers and subscribers. However, to fulfil the said promise, there is a need to provide effective Call Admission Control (CAC) based QoS provisioning solutions which are autonomous, intelligent and scalable.

The importance of providing guaranteed Quality of Service (QoS) cannot be overemphasised, especially in the NGN environment which supports converged services on a common IP transport network. Call Admission Control (CAC) mechanisms do provide QoS to class-based services in a proactive manner. However, due to the factors of complexity, scale and dynamicity of NGN, Machine Learning techniques are favoured to analytical approaches for providing autonomous CAC. This paper is an effort to compare the performance of two such approaches - Neural Networks (NN) and Bayesian Networks (BN), to model the network behaviour and to estimate QoS metrics to be used in the CAC algorithm. It provides a way to find the optimum model training size for accurate predictions. Performance comparison is based on a wide range of experiments through a simulated network in Opnet. The outcome of this comparative study provides some interesting insights into the behaviour of NN and BN models and how they can be utilised for better CAC implementations.

Neural networks and fuzzy systems are both part of what we call soft computing or computational intelligence. Both approaches can be applied to similar classes of problems. Although fuzzy approaches have been successful in control applications in the 1990s this success has not noticeably spread to other domains. Neural networks in contrast enjoy a more steady and widespread commercial success be it in credit card fraud detection or as modules in almost every large data mining software system. Fuzzy systems have not yet enjoyed widespread success in the domain of business applications. It seems software vendors lack incentive for implementing fuzzy software because the benefits have not been made clear by the research community and because fuzzy software is more difficult to implement and use than, for example, neural networks. In order for fuzzy systems to succeed in business applications I suggest to look into building a fuzzy tool kit in a community process to provide an open source reference implementation. This paper looks at some of the key considerations that are important for building a fuzzy tool kit that can support the take-up of fuzzy systems in business applications.

Learning a fuzzy classifier from data is a well-known technique in fuzzy data analysis and many learning algorithms have been proposed, typically in the area of neuro-fuzzy systems. All learning algorithms require a number of parameters to be set by the user. These are typically initial fuzzy partitions for all variables and sometimes also the number of fuzzy rules. Especially, for neuro-fuzzy algorithms the initial choice of parameters can be crucial and if ill-chosen may lead to failure of the learning algorithm. Recent trends in data analysis show that automation is an important issue because it helps to provide advanced analytics to users who are no data analysis experts. In order to fully automate a learning algorithm for fuzzy classifiers we preferably need an algorithm that can determine a suitable initial fuzzy partition for the learning algorithm to start with. In this paper we propose such an algorithm that we have implemented to extend the neuro-fuzzy NEFCLASS. NEFCLASS has recently been integrated into an automatic soft computing platform for intelligent data analysis (SPIDA).

Data analysis tools are still very much a collection of data analysis methods that require analysis experts as users. On the other hand, many business users are keen to apply data analysis to business data in order to understand it or to make predictions to improve on their business decisions. In order to make state-of-the-art data analysis techniques available to such non-experts, we developed a wizard for our data analysis tool SPIDA that selects appropriate data analysis methods given soft high-level requirements. The wizard also configures and runs the chosen methods automatically. This paper describes the general approach to automating data analysis, in particular, the theoretical underpinnings of how to select an appropriate data analysis algorithm

Methods for association rule discovery and pruning assume implicitly that the associations hidden in the data are stable over time and thus provide a rather static view on data and their underlying structure. This is unrealistic in time-stamped domains, which are standard for real life business data. The question "which association rules exist?" is replaced by "how do properties of association rules change?" In order to cope with the vast number of detectable rule changes, preprocessing techniques are required that find those rules which are root cause to interesting rule changes. The paper proposes an approach based on statistical tests that finds derivative rule change histories and marks the respective rules as redundant. The effectiveness in reducing the number of rule histories is demonstrated using real life survey data.

Although fuzzy data analysis has increased in popularity within the research community, this technology is rarely found in industrial solutions. In contrast to fuzzy controllers that were rapidly picked up in the 1990's this has so far not happened for fuzzy data analysis or fuzzy knowledge-based approaches. We believe this is mainly due to not enough easy to use software being available. Software manufactures in the area of data analysis concentrate on statistical approaches and well-known machine learning approaches like decision trees. Typically, neural networks are the only soft computing technique that is sometimes provided. In order to push fuzzy systems and related technology into industrial data analysis applications we need to provide appropriate software. We have developed an IDA platform that automates the data analysis process to a large extent. It uses fuzzy knowledge bases to match user requirements to features of analysis methods and to select, configure and execute IDA processes automatically. Although the platform can use any type of data analysis method we have focused on soft computing methods.

Tele-care systems aim to carry out intelligent analyses of a person's wellbeing using data about their daily activities. This is a very challenging task because the massive dataset is likely to be erroneous, possibly with misleading sections due to noise or missing values. Furthermore, the interpretation of the data is highly sensitive to the lifestyle of the monitored person and the environment in which they interact. In our tele-care project, sensor-network domain knowledge is used to overcome the difficulties of monitoring long-term wellbeing with an imperfect data source. In addition, a fuzzy association analysis is leveraged to implement a dynamic and flexible analysis over individual- and environment-dependent data.

The "unique selling point" of fuzzy systems is usually the interpretability of its rule base. However, very often only the accuracy of the rule base is measured and used to compare a fuzzy system to other solutions. We have suggested an index to measure the interpretability of fuzzy rule bases for classification problems. However, the index can be used to describe the interpretability of any rule-based system that uses sets to partition variables. We demonstrate the features of the index by using two data sets, one simple benchmark set and a real-world example.

In order to use fuzzy systems for knowledge discovery, we need algorithms to induce comprehensible fuzzy systems from data. The comprehensibility of a fuzzy model is mainly determined by its number of rules and variables, but also by meaningful membership functions. We discuss the value of interpretable fuzzy models for intelligent data analysis and highlight some features of the free neuro-fuzzy software NEFCLASS (NEuro-Fuzzy CLASSification) that uses special techniques to induce interpretable fuzzy classifiers from data.