Abstract Through the quantification of physical activity energy expenditure (PAEE), health care monitoring has the potential to stimulate vital and healthy ageing, inducing behavioural changes in older people and linking these to personal health gains. To be able to measure PAEE in a health care perspective, methods from wearable accelerometers have been developed, however, mainly targeted towards younger people. Since elderly subjects differ in energy requirements and range of physical activities, the current models may not be suitable for estimating PAEE among the elderly. Furthermore, currently available methods seem to be either simple but non-generalizable or require elaborate (manual) feature construction steps. Because past activities influence present PAEE, we propose a modeling approach known for its ability to model sequential data, the recurrent neural network (RNN). To train the RNN for an elderly population, we used the growing old together validation (GOTOV) dataset with 34 healthy participants of 60 years and older (mean 65 years old), performing 16 different activities. We used accelerometers placed on wrist and ankle, and measurements of energy counts by means of indirect calorimetry. After optimization, we propose an architecture consisting of an RNN with 3 GRU layers and a feedforward network combining both accelerometer and participant-level data. Our efforts included switching mean to standard deviation for down-sampling the input data and combining temporal and static data (person-specific details such as age, weight, BMI). The resulting architecture produces accurate PAEE estimations while decreasing training input and time by a factor of 10. Subsequently, compared to the state-of-the-art, it is capable to integrate longer activity data which lead to more accurate estimations of low intensity activities EE. It can thus be employed to investigate associations of PAEE with vitality parameters of older people related to metabolic and cognitive health and mental well-being.

Abstract Subgroup discovery (SD) is an exploratory pattern mining paradigm that comes into its own when dealing with large real-world data, which typically involves many attributes, of a mixture of data types. Essential is the ability to deal with numeric attributes, whether they concern the target (a regression setting) or the description attributes (by which subgroups are identified). Various specific algorithms have been proposed in the literature for both cases, but a systematic review of the available options is missing. This paper presents a generic framework that can be instantiated in various ways in order to create different strategies for dealing with numeric data. The bulk of the work in this paper describes an experimental comparison of a considerable range of numeric strategies in SD, where these strategies are organised according to four central dimensions. These experiments are furthermore repeated for both the classification task (target is nominal) and regression task (target is numeric), and the strategies are compared based on the quality of the top subgroup, and the quality and redundancy of the top-k result set. Results of three search strategies are compared: traditional beam search, complete search, and a variant of diverse subgroup set discovery called cover-based subgroup selection. Although there are various subtleties in the outcome of the experiments, the following general conclusions can be drawn: it is often best to determine numeric thresholds dynamically (locally), in a fine-grained manner, with binary splits, while considering multiple candidate thresholds per attribute.

Subgroup Discovery is a supervised, exploratory data mining paradigm that aims to identify subsets of a dataset that show interesting behaviour with respect to some designated target attribute. The way in which such distributional differences are quantified varies with the target attribute type. This work concerns continuous targets, which are important in many practical applications. For such targets, differences are often quantified using z-score and similar measures that compare simple statistics such as the mean and variance of the subset and the data. However, most distributions are not fully determined by their mean and variance alone. As a result, measures of distributional difference solely based on such simple statistics will miss potentially interesting subgroups. This work proposes methods to recognise distributional differences in a much broader sense. To this end, density estimation is performed using histogram and kernel density estimation techniques. In the spirit of Exceptional Model Mining, the proposed methods are extended to deal with multiple continuous target attributes, such that comparisons are not restricted to univariate distributions, but are available for joint distributions of any dimensionality. The methods can be incorporated easily into existing Subgroup Discovery frameworks, so no new frameworks are developed.

Abstract A population group that is often overlooked in the recent revolution of self-tracking is the group of older people. This growing proportion of the general population is often faced with increasing health issues and discomfort. In order to come up with lifestyle advice towards the elderly, we need the ability to quantify their lifestyle, before and after an intervention. This research focuses on the task of activity recognition (AR) from accelerometer data. With that aim, we collect a substantial labelled dataset of older individuals wearing multiple devices simultaneously and performing a strict protocol of 16 activities (the GOTOV dataset, $$N=28$$ N=28 ). Using this dataset, we trained Random Forest AR models, under varying sensor set-ups and levels of activity description granularity. The model that combines ankle and wrist accelerometers (GENEActiv) produced the best results (accuracy $$>80\%$$ >80% ) for 16-class classification. At the same time, when additional physiological information is used, the accuracy increased ( $$>85\%$$ >85% ). To further investigate the role of granularity in our predictions, we developed the LARA algorithm, which uses a hierarchical ontology that captures prior biological knowledge to increase or decrease the level of activity granularity (merge classes). As a result, a 12-class model in which the different paces of walking were merged showed a performance above $$93\%$$ 93% . Testing this 12-class model in labelled free-living pilot data, the mean balanced accuracy appeared to be reasonably high, while using the LARA algorithm, we show that a 7-class model (lying down, sitting, standing, household, walking, cycling, jumping) was optimal for accuracy and granularity. Finally, we demonstrate the use of the latter model in unlabelled free-living data from a larger lifestyle intervention study. In this paper, we make the validation data as well as the derived prediction models available to the community.

Exceptional preferences mining (EPM) is a crossover between two subfields of data mining: local pattern mining and preference learning. EPM can be seen as a local pattern mining task that finds subsets of observations where some preference relations between labels significantly deviate from the norm. It is a variant of subgroup discovery, with rankings of labels as the target concept. We employ several quality measures that highlight subgroups featuring exceptional preferences, where the focus of what constitutes ‘exceptional’ varies with the quality measure: two measures look for exceptional overall ranking behavior, one measure indicates whether a particular label stands out from the rest, and a fourth measure highlights subgroups with unusual pairwise label ranking behavior. We explore a few datasets and compare with existing techniques. The results confirm that the new task EPM can deliver interesting knowledge.

In elite sports, training schedules are becoming increasingly complex, and a large number of parameters of such schedules need to be tuned to the specific physique of a given athlete. In this paper, we describe how extensive analysis of historical data can help optimise these parameters, and how possible pitfalls of under- and overtraining in the past can be avoided in future schedules. We treat the series of exercises an athlete undergoes as a discrete sequence of attributed events, that can be aggregated in various ways, to capture the many ways in which an athlete can prepare for an important test event. We report on a cooperation with the elite speed skating team LottoNL-Jumbo, who have recorded detailed training data over the last 15 years. The aim of the project was to analyse this potential source of knowledge, and extract actionable and interpretable patterns that can provide input to future improvements in training. We present two alternative techniques to aggregate sequences of exercises into a combined, long-term training effect, one of which based on a sliding window, and one based on a physiological model of how the body responds to exercise. Next, we use both linear modelling and Subgroup Discovery to extract meaningful models of the data.

Abstract The problem of Label Ranking is receiving increasing attention from several research communities. The algorithms that have been developed/adapted to treat rankings of a fixed set of labels as the target object, including several different types of decision trees (DT). One DT‐based algorithm, which has been very successful in other tasks but which has not been adapted for label ranking is the Random Forests (RF) algorithm. RFs are an ensemble learning method that combines different trees obtained using different randomization techniques. In this work, we propose an ensemble of decision trees for Label Ranking, based on Random Forests, which we refer to as Label Ranking Forests (LRF). Two different algorithms that learn DT for label ranking are used to obtain the trees. We then compare and discuss the results of LRF with standalone decision tree approaches. The results indicate that the method is highly competitive.

Predefined pattern detection from time series is an interesting and challenging task. In order to reduce its computational cost and increase effectiveness, a number of time series representation methods and similarity measures have been proposed. Most of the existing methods focus on full sequence matching, that is, sequences with clearly defined beginnings and endings, where all data points contribute to the match. These methods, however, do not account for temporal and magnitude deformations in the data and result to be ineffective on several real-world scenarios where noise and external phenomena introduce diversity in the class of patterns to be matched. In this paper, we present a novel pattern detection method, which is based on the notions of templates , landmarks, constraints and trust regions. We employ the Minimum Description Length (MDL) principle for time series preprocessing step , which helps to preserve all the prominent features and prevents the template from overfitting. Templates are provided by common users or domain experts, and represent interesting patterns we want to detect from time series. Instead of utilising templates to match all the potential subsequences in the time series, we translate the time series and templates into landmark sequences, and detect patterns from landmark sequence of the time series. Through defining constraints within the template landmark sequence, we effectively extract all the landmark subsequences from the time series landmark sequence, and obtain a number of landmark segments (time series subsequences or instances). We model each landmark segment through scaling the template in both temporal and magnitude dimensions. To suppress the influence of noise, we introduce the concept of trust region , which not only helps to achieve an improved instance model, but also helps to catch the accurate boundaries of instances of the given template. Based on the similarities derived from instance models, we introduce the probability density function to calculate a similarity threshold. The threshold can be used to judge if a landmark segment is a true instance of the given template or not. To evaluate the effectiveness and efficiency of the proposed method, we apply it to two real-world datasets. The results show that our method is capable of detecting patterns of temporal and magnitude deformations with competitive performance.

Fitting sensors to humans and physical structures is becoming more and more common. These developments provide many opportunities for ubiquitous computing, as well as challenges for analyzing the resulting sensor data. From these challenges, an underappreciated problem arises: modeling multivariate time series with mixed sampling rates. Although mentioned in several application papers using sensor systems, this problem has been left almost unexplored, often hidden in a preprocessing step or solved manually as a one-pass procedure (feature extraction/construction). This leaves an opportunity to formalize and develop methods that address mixed sampling rates in an automatic fashion.

Subgroup Discovery (SD) aims to find coherent, easy-to-interpret subsets of the dataset at hand, where something exceptional is going on. Since the resulting subgroups are defined in terms of conditions on attributes of the dataset, this data mining task is ideally suited to be used by non-expert analysts. The typical SD approach uses a heuristic beam search, involving parameters that strongly influence the outcome. Unfortunately, these parameters are often hard to set properly for someone who is not a data mining expert; correct settings depend on properties of the dataset, and on the resulting search landscape. To remove this potential obstacle for casual SD users, we introduce ROCsearch, a new ROC-based beam search variant for Subgroup Discovery. On each search level of the beam search, ROCsearch analyzes the intermediate results in ROC space to automatically determine a sensible search width for the next search level. Thus, beam search parameter setting is taken out of the domain expert's hands, lowering the threshold for using Subgroup Discovery. Also, ROCsearch automatically adapts its search behavior to the properties and resulting search landscape of the dataset at hand. Aside form these advantages, we also show that ROCsearch is an order of magnitude more efficient than traditional beam search, while its results are equivalent and on large datasets even better than traditional beam search results.

This work presents data mining methods for discovering unusual consumption patterns and their associated descriptive models from smart electricity meter data. At present, data mining and knowledge discovery in electricity meter data suffer from three notable weaknesses: 1) insufficient focus on intelligent data analysis of subgroups (subsets) whose patterns vary significantly from aggregate patterns embodied in an entire dataset; 2) a lack of effort towards generating intuitively understandable and practically applicable knowledge for industrial practitioners to identify such subgroups; and 3) limited knowledge regarding the link between unusual consumption patterns and household consumers' socio-demographic characteristics. This paper addresses these practically important but technically challenging issues by applying subgroup discovery algorithms to a real smart electricity meter dataset. Subgroups whose patterns are unusual and whose sizes are large enough are discovered, and their descriptive and predictive models are generated. Furthermore, to enrich subgroup discovery algorithms, three new-quality measures for real-valued targets are proposed. The comparative studies empirically evaluate the effectiveness and usefulness of subgroup discovery on classification accuracy, predictive power, and computational resources. The methodologies and algorithms presented are generic, and therefore applicable to a wider range of data mining problems.

More and more, physical systems are being fitted with various kinds of sensors in order to monitor their behavior, health or intensity of use. The large quantities of time series data collected from these complex systems often exhibit two important characteristics: the data is a combination of various superimposed effects operating at different time scales, and each effect shows a fair degree of repetition. Each of these effects can be described by a small collection of motifs: recurring temporal patterns in the data. We propose a method to discover characteristic and potentially overlapping motifs at multiple time scales, taking into account systemic deformations and temporal warping. Our method is based on a combination of scale-space theory and the Minimum Description Length principle. We show its effectiveness on two time series datasets from real world applications.

Subgroup discovery systems are concerned with finding interesting patterns in labeled data. How these systems deal with numeric and nominal data has a large impact on the quality of their results. In this paper, we consider two ways to extend the standard pattern language of subgroup discovery: using conditions that test for interval membership for numeric attributes, and value set membership for nominal attributes. We assume a greedy search setting, that is, iteratively refining a given subgroup, with respect to a (convex) quality measure. For numeric attributes, we propose an algorithm that finds the optimal interval in linear (rather than quadratic) time, with respect to the number of examples and split points. Similarly, for nominal attributes, we show that finding the optimal set of values can be achieved in linear (rather than exponential) time, with respect to the number of examples and the size of the domain of the attribute. These algorithms operate by only considering subgroup refinements that lie on a convex hull in ROC space, thus significantly narrowing down the search space. We further provide efficient algorithms specifically for the popular Weighted Relative Accuracy quality measure, taking advantage of some of its properties. Our algorithms are shown to perform well in practice, and furthermore provide additional expressive power leading to higher-quality results.

Exceptional Model Mining (EMM) is an exploratory data analysis technique that can be regarded as a generalization of subgroup discovery. In EMM we look for subgroups of the data for which a model fitted to the subgroup differs substantially from the same model fitted to the entire dataset. In this paper we develop methods to mine for exceptional regression models. We propose a measure for the exceptionality of regression models (Cook's distance), and explore the possibilities to avoid having to fit the regression model to each candidate subgroup. The algorithm is evaluated on a number of real life datasets. These datasets are also used to illustrate the results of the algorithm. We find interesting subgroups with deviating models on datasets from several different domains. We also show that under certain circumstances one can forego fitting regression models on up to 40% of the subgroups, and these 40% are the relatively expensive regression models to compute.

Large data is challenging for most existing discovery algorithms, for several reasons. First of all, such data leads to enormous hypothesis spaces, making exhaustive search infeasible. Second, many variants of essentially the same pattern exist, due to (numeric) attributes of high cardinality, correlated attributes, and so on. This causes top- k mining algorithms to return highly redundant result sets, while ignoring many potentially interesting results. These problems are particularly apparent with subgroup discovery (SD) and its generalisation, exceptional model mining. To address this, we introduce subgroup set discovery : one should not consider individual subgroups, but sets of subgroups. We consider three degrees of redundancy, and propose corresponding heuristic selection strategies in order to eliminate redundancy. By incorporating these (generic) subgroup selection methods in a beam search, the aim is to improve the balance between exploration and exploitation. The proposed algorithm, dubbed DSSD for diverse subgroup set discovery, is experimentally evaluated and compared to existing approaches. For this, a variety of target types with corresponding datasets and quality measures is used. The subgroup sets that are discovered by the competing methods are evaluated primarily on the following three criteria: (1) diversity in the subgroup covers (exploration), (2) the maximum quality found (exploitation), and (3) runtime. The results show that DSSD outperforms each traditional SD method on all or a (non-empty) subset of these criteria, depending on the specific setting. The more complex the task, the larger the benefit of using our diverse heuristic search turns out to be.

Subgroup discovery suffers from the multiple comparisons problem: we search through a large space, hence whenever we report a set of discoveries, this set will generally contain false discoveries. We propose a method to compare subgroups found through subgroup discovery with a statistical model we build for these false discoveries. We determine how much the subgroups we find deviate from the model, and hence statistically validate the found subgroups. Furthermore we propose to use this subgroup validation to objectively compare quality measures used in subgroup discovery, by determining how much the top subgroups we find with each measure deviate from the statistical model generated with that measure. We thus aim to determine how good individual measures are in selecting significant findings. We invoke our method to experimentally compare popular quality measures in several subgroup discovery settings.

Whenever a dataset has multiple discrete target variables, we want our algorithms to consider not only the variables themselves, but also the interdependencies between them. We propose to use these interdependencies to quantify the quality of subgroups, by integrating Bayesian networks with the Exceptional Model Mining framework. Within this framework, candidate subgroups are generated. For each candidate, we fit a Bayesian network on the target variables. Then we compare the network's structure to the structure of the Bayesian network fitted on the whole dataset. To perform this comparison, we define an edit distance-based distance metric that is appropriate for Bayesian networks. We show interesting subgroups that we experimentally found with our method on datasets from music theory, semantic scene classification, biology and zoogeography.

Overthelastdecade,localpatterndiscoveryhasbecomearapidlygrowingfield(Moriketal.2005),andarangeoftechniquesisavailableforproducingextensivecollectionsofpatterns.Becauseoftheexhaustivenatureofmostsuchtechniques,thepatterncol-lections provide a fairly complete picture of the information content of the database.However,suchso-calledlocalpatternsrepresentfragmentedknowledge,anditisoftennot clear how the pieces of the puzzle can be combined into a global model, which isoften the desirable result of a data mining process. Thus, the question of how to turnlarge collections of patterns into global models deserves attention.This special issue of the Data Mining and Knowledge Discovery Journal featuresa number of papers that represent the state of the art in building global models fromlocal patterns. In our view, a common ground of all the local pattern mining tech-niquesisthattheycanbeconsidered tobefeatureconstructiontechniques thatfollowdifferent objectives (or constraints). We will see that the redundancy of these patternsandtheselectionofsuitablesubsetsofpatternsareaddressedinseparatesteps,sothateach resulting feature is highly informative in the context of the global data miningproblem.In earlier work (Knobbe et al. 2008), a framework was proposed that provides ageneral outline of the activities involved. The framework, called From Local Patterns

Spatial classification is the task of learning models to predict class labels based on the features of entities as well as the spatial relationships to other entities and their features. Spatial data can be represented as multi-relational data, however it presents novel challenges not present in multi-relational problems. One such problem is that spatial relationships are embedded in space, unknown a priori, and it is part of the algorithm’s task to determine which relationships are important and what properties to consider. In order to determine when two entities are spatially related in an adaptive and non-parametric way, we propose a Voronoi-based neighbourhood definition upon which spatial literals can be built. Properties of these neighbourhoods also need to be described and used for classification purposes. Non-spatial aggregation literals already exist within the multi-relational framework, but are not sufficient for comprehensive spatial classification. A formal set of additions to the multi-relational data mining framework is proposed, to be able to represent spatial aggregations as well as spatial features and literals. These additions allow for capturing more complex interactions and spatial occurrences such as spatial trends. In order to more efficiently perform the rule learning and exploit powerful multi-processor machines, a scalable parallelized method capable of reducing the runtime by several factors is presented. The method is compared against existing methods by experimental evaluation on a real world crime dataset which demonstrate the importance of the neighbourhood definition and the advantages of parallelization.

In this paper we present a new approach to mining binary data. We treat each binary feature (item) as a means of distinguishing two sets of examples. Our interest is in selecting from the total set of items an itemset of specified size, such that the database is partitioned with as uniform a distribution over the parts as possible. To achieve this goal, we propose the use of joint entropy as a quality measure for itemsets, and refer to optimal itemsets of cardinality k as maximally informative k-itemsets. We claim that this approach maximises distinctive power, as well as minimises redundancy within the feature set. A number of algorithms is presented for computing optimal itemsets efficiently.

This paper describes experiments in applying data mining techniques to historical data collected by network monitoring agents. Large amounts of performance data, including network, system, and application performance data, are collected and stored by monitoring agents. Data mining algorithms analyze the data and codify it into usable knowledge. We show, via experiments, that the knowledge contains useful and unexpected suggestions for improving the effectiveness of business processes and for reducing management support effort. Four experiments are discussed: three preliminary laboratory experiments and one large, real-world experiment at a major airline company.