Search

[EN] CRISP-DM (CRoss-Industry Standard Process for Data Mining) has its origins in the second half of the nineties and is thus about two decades old. According to many surveys and user polls it is still thede factostandard for developing data mining and knowledge discovery projects. However, undoubtedly the field has moved on considerably in twenty years, with data science now the leading term being favoured over data mining. In this paper we investigate whether, and in what contexts, CRISP-DM is still fit for purpose for data science projects. We argue that if the project is goal-directed and process-driven the process model view still largely holds. On the other hand, when data science projects become more exploratory the paths that the project can take become more varied, and a more flexible model is called for. We suggest what the outlines of such a trajectory-based model might look like and how it can be used to categorise data science projects (goal-directed, exploratory or data management). We examine seven real-life exemplars where exploratory activities play an important role and compare them against 51 use cases extracted from the NIST Big Data Public Working Group. We anticipate this categorisation can help project planning in terms of time and cost characteristics., We thank the anonymous reviewers for their comments, which motivated the analysis in Section 5. This material is based upon work supported by the EU (FEDER), and the Spanish MINECO under Grant RTI2018-094403-B-C3, the Generalitat Valenciana PROMETEO/2019/098. F. MartinezPlumed was also supported by INCIBE (Ayudas para la excelencia de los equipos de investigacion avanzada en ciberseguridad), the European Commission (JRC) HUMAINT project (CT-EX2018D335821-101), and UPV (PAID-06-18). J. H-Orallo is also funded by an FLI grant RFP2-152.

Applications running on HPC systems waste time and energy if they: (a) use resources inefficiently, (b) deviate from allocation purpose (e.g. cryptocurrency mining), or (c) encounter errors and failures. It is important to know which applications are running on the system, how they use the system, and whether they have been executed before. To recognize known applications during execution on a noisy system, we draw inspiration from the way Shazam recognizes known songs playing in a crowded bar. Our contribution is an Execution Fingerprint Dictionary (EFD) that stores execution fingerprints of system metrics (keys) linked to application and input size information (values) as key-value pairs for application recognition. Related work often relies on extensive system monitoring (many system metrics collected over large time windows) and employs machine learning methods to identify applications. Our solution only uses the first 2 minutes and a single system metric to achieve F-scores above 95 percent, providing comparable results to related work but with a fraction of the necessary data and a straightforward mechanism of recognition.

Inductive Logic Programming (ILP) is a field at the intersection of Machine Learning and Logic Programming, based on logic as a uniform representation language for expressing examples, background knowledge and hypotheses.

Companies want to extract value from their relational databases. This is the aim of relational data mining. Propositionalization is one possible approach to relational data mining. Propositionalization adds new attributes, called features, to the main table, leading to an attribute-value representation, a single table, on which a propositional learner can be applied. However, current relational databases are large and composed of mixed, numerical and categorical, data. Moreover, the specificity of relational data is to involve one-to-many relationships. As an example of such data, consider customers purchasing products: each customer can purchase several products. Therefore, there is a need for techniques able to learn complex aggregates. Learning such features means to explore a combinatorial, possibly infinite, space and such an approach is prone to overfitting. We introduce a propositionalization approach dedicated to a robust Bayesian classifier. It efficiently samples a given number of features in the language bias, following a distribution over the complex aggregates. This distribution is also used to penalize complex aggregates in the regularization of the robust Bayesian classifier. Experiments show that it performs better than state-of-the-art methods on most investigated benchmarks and can deal with large datasets more easily. A new real, large, mixed relational dataset is introduced which confirms the ability of our approach to learn complex aggregates.

Vegetation monitoring is becoming a major issue in the urban environment due to the services they procure and necessitates an accurate and up to date mapping. Very High Resolution satellite images enable a detailed mapping of the urban tree and herbaceous vegetation. Several supervised classifications with statistical learning techniques have provided good results for the detection of urban vegetation but necessitate a large amount of training data. In this context, this study proposes to investigate the performances of different sampling strategies in order to reduce the number of examples needed. Two windows based active learning algorithms from state-of-art are compared to a classical stratified random sampling and a third combining active learning and stratified strategies is proposed. The efficiency of these strategies is evaluated on two medium size French cities, Strasbourg and Rennes, associated to different datasets. Results demonstrate that classical stratified random sampling can in some cases be just as effective as active learning methods and that it should be used more frequently to evaluate new active learning methods. Moreover, the active learning strategies proposed in this work enables to reduce the computational runtime by selecting multiple windows at each iteration without increasing the number of windows needed.

Our approach can handle thresholds on attributes and on the number of objects.Tackling numeric attributes with both absolute and relative numbers efficiently.Selecting the optimal combination of propositionalizer and classifier effectively.The proposed approach is flexible to be applied over different contexts.Experiments show the effectiveness and efficiency of the proposed approach. In a relational database, data are stored in primary and secondary tables. Propositionalization can transform a relational database into a single attribute-value table, and hence becomes a useful technique for mining relational databases. However, most of the existing propositionalization approaches deal with categorical attributes, and cannot handle a threshold on an attribute and a threshold on the number of objects satisfying the condition on the attribute at the same time. In this paper, we propose a new propositionalization technique called Cardinalization to solve these problems. In order to handle relative numbers, we propose a second variant of our approach called Quantiles which can discretize the cardinality of Cardinalization and achieve a fixed number of features. Therefore, the Quantiles method can be tuned to different deployment contexts. Additionally, we often observe that the best combination of propositionalization and classification methods depends on the new context (e.g., online/incremental learning). One effective solution could be to predict the optimal combination at training time and use it in different deployment contexts. Here we also propose an effective wrapping algorithm, called WPC (Wrapper to combine Propositionalizer and Classifier) to select the best combination of propositionalization and classification methods to address this task. Extensive performance analyses in synthetic and real-life datasets show that our approach is very effective and efficient in relational data mining.

Propositionalization is the process of explicitly transforming a relational dataset into a propositional dataset.

Active learning (AL) algorithms have been proven useful in reducing the number of required training samples for remote sensing applications; however, most methods query samples pointwise without considering spatial constraints on their distribution. This may often lead to a spatially dispersed distribution of training points unfavorable for visual image interpretation or field surveys. The aim of this study is to develop region-based AL heuristics to guide user attention toward a limited number of compact spatial batches rather than distributed points. The proposed query functions are based on a tree ensemble classifier and combine criteria of sample uncertainty and diversity to select regions of interest. Class imbalance, which is inherent to many remote sensing applications, is addressed through stratified bootstrap sampling. Empirical tests of the proposed methods are performed with multitemporal and multisensor satellite images capturing, in particular, sites recently affected by large-scale landslide events. The assessment includes an experimental evaluation of the labeling time required by the user and the computational runtime, and a sensitivity analysis of the main algorithm parameters. Region-based heuristics that consider sample uncertainty and diversity are found to outperform pointwise sampling and region-based methods that consider only uncertainty. Reference landslide inventories from five different experts enable a detailed assessment of the spatial distribution of remaining errors and the uncertainty of the reference data.

Correlation mining is recognized as one of the most important data mining tasks for its capability to identify underlying dependencies between objects. On the other hand, graph-based data mining techniques are increasingly applied to handle large datasets due to their capability of modeling various non-traditional domains representing real-life complex scenarios such as social/computer networks, map/spatial databases, chemical-informatics domain, bio-informatics, image processing and machine learning. To extract useful knowledge from large amount of spurious patterns, correlation measures are used. Nonetheless, existing graph based correlation mining approaches are unable to capture effective correlations in graph databases. Hence, we have concentrated on graph correlation mining and proposed a new graph correlation measure, gConfidence, to discover more useful graph patterns. Moreover, we have developed an efficient algorithm, CGM (Correlated Graph Mining), to find the correlated graphs in graph databases. The performance of our scheme was extensively analyzed in several real-life and synthetic databases based on runtime and memory consumption, then compared with existing graph correlation mining algorithms, which proved that CGM is scalable with respect to required processing time and memory consumption and outperforms existing approaches by a factor of two in speed of mining correlations.

We describe a systematic approach called reframing, defined as the process of preparing a machine learning model (e.g., a classifier) to perform well over a range of operating contexts. One way to achieve this is by constructing a versatile model, which is not fitted to a particular context, and thus enables model reuse. We formally characterise reframing in terms of a taxonomy of context changes that may be encountered and distinguish it from model retraining and revision. We then identify three main kinds of reframing: input reframing, output reframing and structural reframing. We proceed by reviewing areas and problems where some notion of reframing has already been developed and shown useful, if under different names: re-optimising, adapting, tuning, thresholding, etc. This exploration of the landscape of reframing allows us to identify opportunities where reframing might be possible and useful. Finally, we describe related approaches in terms of the problems they address or the kind of solutions they obtain. The paper closes with a re-interpretation of the model development and deployment process with the use of reframing., We thank the anonymous reviewers for their comments, which have helped to improve this paper significantly. This work was supported by the REFRAME project, granted by the European Coordinated Research on Long-term Challenges in Information and Communication Sciences Technologies ERA-Net (CHIST-ERA), funded by their respective national funding agencies in the UK (EPSRC, EP/K018728), France and Spain (MINECO, PCIN-2013-037). It has also been partially supported by the EU (FEDER) and Spanish MINECO grant TIN2015-69175-C4-1-R and by Generalitat Valenciana PROMETEOII/2015/013.

Chemical reactions always involve several molecules of two types, reactants and products. Existing data mining techniques, eg. Quantitative Structure Activity Relationship (QSAR) methods, deal with individual molecules only. In this article, we propose to use a Condensed Graph of Reaction (CGR) to merge all molecules involved in a reaction into one molecular graph. This allows one to consider reactions as pseudo-molecules and to develop QSAR models based on fragment descriptors. Then ISIDA (In SIlico Design and Analysis) fragment descriptors built from CGRs are used to generate models for the rate constant of S N 2 reactions in water, using three usual attribute-value regression algorithms (linear regression, support vector machine, and regression trees). This approach is compared favorably to two state-of-the-art relational data mining techniques.

Some supervised tasks are presented with a numerical output but decisions have to be made in a discrete, binarised, way, according to a particular cutoff. This binarised regression task is a very common situation that requires its own analysis, different from regression and classification—and ordinal regression. We first investigate the application cases in terms of the information about the distribution and range of the cutoffs and distinguish six possible scenarios, some of which are more common than others. Next, we study two basic approaches: the retraining approach, which discretises the training set whenever the cutoff is available and learns a new classifier from it, and the reframing approach, which learns a regression model and sets the cutoff when this is available during deployment. In order to assess the binarised regression task, we introduce context plots featuring error against cutoff. Two special cases are of interest, the UCEUCE and OCEOCE curves, showing that the area under the former is the mean absolute error and the latter is a new metric that is in between a ranking measure and a residual-based measure. A comprehensive evaluation of the retraining and reframing approaches is performed using a repository of binarised regression problems created on purpose, concluding that no method is clearly better than the other, except when the size of the training data is small., We thank the anonymous reviewers for their comments, which have helped to improve this paper significantly. We thank Peter Flach and Meelis Kull for their insightful comments and very useful suggestions. This work was supported by the Spanish MINECO under Grant TIN 2013-45732-C4-1-P and by Generalitat Valenciana PROMETEOII2015/013. This research has been developed within the REFRAME project, granted by the European Coordinated Research on Long-term Challenges in Information and Communication Sciences & Technologies ERA-Net (CHIST-ERA), and funded by the Ministerio de Economia y Competitividad in Spain (PCIN-2013-037) and the Agence Nationale pour la Recherche in France (ANR-12-CHRI-0005-03).

The first international workshop on Learning over Multiple Contexts, devoted to generalization and reuse of machine learning models over multiple contexts, was held on September 19th, 2014, as part of the 7th European machine learning and data mining conference (ECML-PKDD 2014) in Nancy, France. This short report summarizes the presentations and discussions held during the LMCE 2014 workshop, as well as the workshop conclusions and the future agenda.

In this article we consider the application of the Transductive Ridge Regression (TRR) approach to structure-activity modeling. An original procedure of the TRR parameters optimization is suggested. Calculations performed on 3 different datasets involving two types of descriptors demonstrated that TRR outperforms its non-transductive analogue (Ridge Regression) in more than 90 % of cases. The most significant transductive effect was observed for small datasets. This suggests that transduction may be particularly useful when the data are expensive or difficult to collect.

This paper deals with learning first-order logic rules from data lacking an explicit classification predicate. Consequently, the learned rules are not restricted to predicate definitions as in supervised inductive logic programming. First-order logic offers the ability to deal with structured, multi-relational knowledge. Possible applications include first-order knowledge discovery, induction of integrity constraints in databases, multiple predicate learning, and learning mixed theories of predicate definitions and integrity constraints. One of the contributions of our work is a heuristic measure of confirmation, trading off novelty and satisfaction of the rule. The approach has been implemented in the Tertius system. The system performs an optimal best-first search, finding the k most confirmed hypotheses, and includes a non-redundant refinement operator to avoid duplicates in the search. Tertius can be adapted to many different domains by tuning its parameters, and it can deal either with individual-based representations by upgrading propositional representations to first-order, or with general logical rules. We describe a number of experiments demonstrating the feasibility and flexibility of our approach.

In the next few decades the increase of the number of elderly people will be of major concern, so that solutions must be found in order to maintain them at home. However such a population is exposed to the risk of falls, that can lead to dependency. This paper recalls some approaches used for fall detection and focuses on a method based on an uncalibrated camera. Motion detection uses a combination of simple Gaussian background modelling and interframe difference for person shape detection and features extraction. These features feed a Hidden Markov Model dedicated to fall detection. The algorithm has been tested on real data and we show that simple techniques can be used in order to obtain a fast and reliable fall detection system.

International audience; The aim of our research is to analyze the evolution of urbanization and to simulate it on specific areas. We focus on the evolution between 1950 and now. We analyse the densification by means of comparing temporal topographic data bases created from existing topographic data base and maps and photo from 1950. In this paper we present how a simulation works - which input data are used, which functions are used to densify the space and how the simulation works, is tuned and run - the densification method for each urban block illustrated with results, the method used during the project to build the required knowledge for simulation and we conclude and present the main research perspectives. The methods are implemented on a dedicated open source software named GeOpenSim.

A policy gradient method is a reinforcement learning approach that directly optimizes a parametrized control policy by a variant of gradient descent. These methods belong to the class of policy search techniques that maximize the expected return of a policy in a fixed policy class, in contrast with traditional value function approximation approaches that derive policies from a value function. Policy gradient approaches have various advantages: they enable the straightforward incorporation of domain knowledge in policy parametrization and often an optimal policy is more compactly represented than the corresponding value function; many such methods guarantee to convergence to at least a locally optimal policy; the methods naturally handle continuous states and actions and often even imperfect state information. The counterveiling drawbacks include difficulties in off-policy settings, the potential for very slow convergence and high sample complexity, as well as identifying local optima that are not globally optimal.

"Chimiotheque Nationale" (CN) represents a library of synthetic and natural products from various French public laboratories. Recently, experimental screening for S. Aureus antibacterial activity of the part of this library has been performed by Dr J.-M. Paris and collaborators in Ecole des Mines (Paris, France). Here, experimental results of the screening have been used to build classification models using SMF descriptors and ISIDA modeling tools (Naive Bayes (NB) and SVM modules). The dataset consisted in a large and structurally diverse set of 4563 compounds, 62 of which having a demonstrated antibacterial activity. In NB calculations, several variations of the algorithm were used. In particular, the a priori distribution of SMF descriptors was modeled using a binomial law, Bernouilli distributions or first order logic. SVM calculations were performed with an RBF kernel which parameters (C,γ) were optimized to maximize the balanced accuracy. Both NB and SVM models were validated using external 5-fold cross validation, repeated three times on randomized data set. Additionally fifteen Y-randomizations were performed in order to check for chance correlation. Predictive performance of the models has been assessed by combination of Precision and Recall parameters: the models with Precision > 0.1 and Recall > 0.6 correspond to over 10-fold enrichment and, therefore, were considered as acceptable. A consensus model was applied to estimate the antibacterial activity of 122 new compounds for the CN. All predictions were done blindly.

Functional Magnetic Resonance Imaging (fMRI) allows the neuroscientists to observe the human brain in vivo. The current approach consists in statistically validating their hypotheses. Data mining techniques provide an opportunity to help them in making up their hypotheses. This paper shows how a neuronal clustering technique can highlight active areas thanks to an appropriate distance between fMRI image sequences. This approach has been integrated into an interactive environment for knowledge discovery in brain fMRI. Its results on a typical dataset validate the approach and open further developments in this direction.

Scope classification is a new instance-based learning (IBL) technique with a rule-based characterisation. Within the scope approach, the classification of an object o is based on the examples that are closer to o than every example labelled with another class. In contrast to standard distance-based IBL classifiers, scope classification relies on partial preorderings ≤o between examples, indexed by objects. Interestingly, the notion of closeness to o that is used characterises the classes predicted by all the rules that cover o and are relevant and consistent for the training set. Accordingly, scope classification is an IBL technique with a rule-based characterisation. Since rules do not have to be explicitly generated, the scope approach applies to classification problems where the number of rules prevents them from being exhaustively computed. Currently at Department of Computer Science, University of Bristol, Merchant Venturers Building, Woodland Road, Bristol BS8 1UB, United Kingdom

Confirmatory induction is based on the assumption that unknown individuals are similar to known ones, i.e. they satisfy the properties shared by known individuals. This assumption can be represented inside a non-monotonic logical framework. Accordingly, existing approaches to confirmatory induction take advantage of the machinery developed so far for non-monotonic inference. However, they are based on completion policies that are unnecessary strong for the induction purpose. The contribution of this paper is twofold: some basic requirements that any model for generalization based on confirmatory induction should satisfy are proposed. Then, a model for generalization based on Hempel's notion of confirmation is introduced. This model is rational in the sense that it satisfies the rationality postulates we exhibit; moreover, the completion principle on which this model is based captures exactly the similarity assumption, hence the model can be considered minimal as well.

The industry resorts to numerous information mapping, text mining and data mining technologies to tap into the patent information goldmine while academic researchers often content themselves with geographically limited information and familiar analysis tools. In this communication, we shall review as comprehensively as possible the practice of both domains and show how they could be brought closer to each other., L’information brevet est souvent perçue comme une mine d’or pour l’industrie et comme un puits sans fond pour la recherche académique. La demande de l’industrie pour des outils lui permettant d’exploiter cette mine d’or en ont fait un utilisateur routinier de nombreuses technologies de cartographie de l’information, de fouille de texte ou de données, quand les chercheurs se contentent souvent d’une information brevet géographiquement limitée et des méthodes d’analyse qui leur sont familières. Malgré leurs différences d’approches et d’objectifs, les deux domaines peuvent apprendre l’un de l’autre. Dans cette communication, nous tenterons de rendre compte, de manière aussi exhaustive que possible, des pratiques qui ont cours dans l’industrie ou dans la recherche et de montrer comment certains dispositifs de navigation entre niveaux méso- et macroscopique pourraient les rapprocher.