Dynamic classifier selection (DCS) regards well-known machine learning techniques in the batch setting that leverage ensemble performance. Most of the methods use similarity-based methods as a proxy, culminating in high computation costs and becoming unfeasible in many streaming scenarios. In this paper, we propose a DCS method able to cope with the high-speed streaming setting, which is based on the performance of base learners in the most recent instances. The impact of our method is evaluated with different ensembles for data streams. We also propose modifications to an Online Boosting method, which has its performance improved with DCS. Our method increases the accuracy and kappa statistic of state-of-the-art ensembles with low overhead of time processing and memory.

The approaches that currently constitute the state-of-the-art for the task of regression on continuous data streams usually involve ensembles, regression trees, and regression rules. They have been found to work very well for certain situations but generally consume computational resources to a prohibitive extent. In this paper, we propose a new method based on an ensemble of linear regressions for the regression task adapted to handle continuous data streams. The technique has been named Adaptive Linear Regression (ALR). The algorithm combines strategies that contribute to high prediction accuracy using (i) distinct sliding window sizes for training each ensemble element, and (ii) a dynamic regressor selection method for final ensemble voting. After an extensive experimental study, ALR was found to exhibit high predictive performance and outperform state-of-the-art ensemble regressors on data streams for real and synthetic datasets. Moreover, it exhibits low processing time in its parallel version and is faster than ARF-Reg in its serial version. The paper also presents an analysis of how the choice of sliding window size for training favors accuracy.

Personality Computing is a field related to artificial intelligence that can recognize personality automatically from texts (TB-APR). To accomplish this task, personality inventories are used to build a corpus for supervised machine learning. Although personality inventories are easy options for responding to and rating personality, they lack efficient mechanisms to control omissions (proposital or non-conscient) and adulterations of personality characteristics. In our literature review, we found that state-of-art models have low to moderate correlations between texts and personality scales; we think the flaws of personality inventories can explain that. The present study provides a new TB-APR approach, using a projective technique to build corpus, for avoiding personality inventories’ biases. For this task, we chose the Thematic Apperception Test, rated by the Revised Morvalian System (RMS), producing an appropriate quantity of texts during the application time. The RMS, which is a revision of a qualitative system developed in the 1980’s, provides not yet explored personality dimensions for artificial intelligence and TB-APR models. The proposed model used the TF-IDF technique with some state-of-art classification algorithms. The results are promising, with an AUC-ROC of average 0.853, suggesting strong correlations between the texts and the scales proposed by the RMS.

The financial credibility of a person is a relevant factor to determine whether a loan should be approved or not, and it is quantified by a credit score, which is computed using past performance on debt obligations, profiling, and other data available. Credit scoring becomes even a hotter topic in emerging countries, as interest rates and customer behavior swiftly vary, given the economic (in)stability of the country and as fintechs are chasing robust solutions for improved credit scoring solutions. Batch machine learning is often deployed for credit scoring, yet, they are tailored for static scenarios, i.e., they are not prepared to swiftly detect and adapt to changes in customer behavior, thus leading to slow recovery in such scenarios. In this paper, we bring forward an analysis on how batch machine learning can be combined with data stream mining techniques, thus leading to better recognition rates in credit scoring scenarios. We analyze three different real-world datasets from Brazilian financial institutions, whilst keeping their secrecy preserved, and show how batch and stream learning can be combined towards improved credit scoring systems, as well as highlighting relevant gaps that still require attention.

This paper provides a new TB-APR approach, using a projective test to build a corpus. The research of Personality Computing shows that it is possible to recognize personality automatically from texts (TB-APR), using paradigms of supervised learning. As concerns these paradigms, texts need to be labeled by psychometric instruments and, in order to realize this task, personality inventories are used. Personality inventories display great facilities for application and correction, but they do not evince efficient ways of controlling intentional or non-conscious omissions of undesired personality characteristics by the individual, which may explain the low correlations found in literature regarding TB-APR models. In this article, we propose the labeling of a textual corpus using the Z-test projective instrument, in order to mitigate the limitations of inventories, since it is very sensitive and offers the possibility of collective application. The proposed model used bag-of-words techniques, with some state of art machine learning inductors. The results are promising, with AUC-ROC on 0.85 average.

Adaptive recommender systems are increasingly showing their importance as profiling is a dynamic problem. Their goal is to update recommendation models as new interactions take place, thus swiftly adapting to drifts in the user's behavior and desires, and item's audience. However, existing recommendation algorithms usually do not perform well during drifts, as they take long to adapt to changes, or these updates are suboptimal since they account for all profiles' preferences equally, which is often untrue as each individual and its changes are unique. In this paper, we propose the ADADRIFT algorithm to deal with user and item-based drifts in adaptive recommender systems using personalized learning rates based on profile statistics. The experiments using stream-based recommender systems (ISGD and BRISMF) across four different datasets show that ADADRIFT surpasses ADADELTA with significant improvements in recommendation rates. The best results appear when the data streams have a long history of the users' or items' interactions and drifts become noticeable. The experimentation in this work highlight the importance of handling drifts in recommender systems.

This paper proposes a hybrid ensemble learning approach that combines statistical and data stream mining algorithms to obtain better forecasting performance in multiple time series prediction problems. Although some multiple time series algorithms perform surprisingly well in a variety of domains, it is well-known that no one is dominant for every existent domain. Therefore, we developed a meta-technique based on data stream mining and static ensemble selection strategy and evaluated its forecasting goodness-of-fit in time series datasets from M3 and M4 competitions. After training different regression models, we show how the combination of auto.arima and AdaGrad leads to improved forecasting rates, thus surpassing the results of state-of-art algorithms.

Abstract Ensemble of classifiers can improve classification accuracy by combining several models. The fusion method plays an important role in the ensemble performance. Usually, a criterion for weighting the decision of each ensemble member is adopted. Frequently, this can be done using some heuristic based on accuracy or confidence. Then, the used fusion rule must consider the established criterion for providing a most reliable ensemble output through a kind of competition among the ensemble members. This article presents a new ensemble fusion method, named centrality score‐based fusion, which uses the centrality concept in the context of social network analysis (SNA) as a criterion for the ensemble decision. Centrality measures have been applied in the SNA to measure the importance of each person inside of a social network, taking into account the relationship of each person with all others. Thus, the idea is to derive the classifier weight considering the overall classifier prominence inside the ensemble network, which reflects the relationships among pairs of classifiers. We hypothesized that the prominent position of a classifier based on its pairwise relationship with the other ensemble members could be its weight in the fusion process. A robust experimental protocol has confirmed that centrality measures represent a promising strategy to weight the classifiers of an ensemble, showing that the proposed fusion method performed well against the literature.

The distributed constraint optimization problem (DCOP) has emerged as one of the most promising coordination techniques in multi-agent systems (MAS). However, because DCOP is known to be NP-hard, the existing DCOP techniques are often unsuitable for large-scale applications, which require scalable algorithms to deal with severely limited computing and communication. Moreover, the selection of DCOP algorithm is a challenging and critical task for obtaining a desirable performance on certain MAS domains. In this paper, we present a performance analysis of incomplete DCOP algorithms on large-scale DCOPs. We experimentally evaluate the state-of-the-art incomplete algorithms on two types of problems involving hundreds of variables with different network topologies and densities. Such performance analysis can help to mitigate the challenges of selection of algorithm for a number of realistic large-scale, complex MAS applications.

The distributed constraint optimization problem (DCOP) has emerged as one of the most promising coordination techniques in multiagent systems. However, because DCOP is known to be NP-hard, the existing DCOP techniques are often unsuitable for large-scale applications, which require distributed and scalable algorithms to deal with severely limited computing and communication. In this paper, we present a novel approach to provide approximate solutions for large-scale, complex DCOPs. This approach introduces concepts of synchronization of coupled oscillators for speeding up the convergence process towards high-quality solutions. We propose a new anytime local search DCOP algorithm, called Coupled Oscillator OPTimization (COOPT), which amounts to iteratively solving a DCOP by agents exchanging local information that brings them to a consensus. We empirically evaluate COOPT on constraint networks involving hundreds of variables with different topologies, domains, and densities. Our experimental results demonstrate that COOPT outperforms other incomplete state-of-the-art DCOP algorithms, especially in terms of the agents' communication cost and solution quality.

This paper presents the Centrality Based Fusion (CBF) method for ensemble fusion which is based on the centrality measures in the context of complex network theory. Such a concept has been applied in Social Network Analysis to measure the importance of each person inside of a social network. We hypothesized that the centrality of each classifier inside of an ensemble represented as a complex network could be combined with accuracy to provide the weight for its decision during the ensemble fusion. The main idea is to derive the weight considering the classifier importance inside the ensemble network which reflects the classifiers' diversity. A robust experimental protocol based on 30 datasets has confirmed that the notion of prominence provided employing centrality measures is a promising strategy to weight the classifiers of an ensemble. When compared with 9 fusion methods of the literature, the proposed fusion method won in 189 out of 270 experiments (70%), lost in 61 cases (22.59%) and tied in 20 cases (7.41%).

The financial market is one of the major consumers of data mining techniques, and the main reason is their efficiency to analyze complex data. One important trait shared between most financial applications is class imbalance. Since traditional classification methods assume nearly balanced classes and equal misclassification costs, they usually fail to deal with imbalanced data. However, in financial contexts, problems are usually imbalanced, and instances from the minority class are known for deficits of millions of dollars every year, e.g., credit card frauds, money laundering transactions and so forth. Over the years, several techniques for dealing with class imbalance have been developed, such as sampling techniques and algorithm adaptations. In this study, we analyze how different sampling techniques impact the performance of different classification systems on financial applications. Results show that, for the given datasets, sampling techniques allow the improvement of prediction performance of the minority class while also improving overall classification rates. Nevertheless, their use often deteriorates the performance in predicting the majority class.

Peer-to-peer (P2P) lending is a global trend of financial markets that allow individuals to obtain and concede loans without having financial institutions as a strong proxy. As many real-world applications, P2P lending presents an imbalanced characteristic, where the number of creditworthy loan requests is much larger than the number of non-creditworthy ones. In this work, we wrangle a real-world P2P lending data set from Lending Club, containing a large amount of data gathered from 2007 up to 2016. We analyze how supervised classification models and techniques to handle class imbalance impact creditworthiness prediction rates. Ensembles, cost-sensitive and sampling methods are combined and evaluated along logistic regression, decision tree, and bayesian learning schemes. Results show that, in average, sampling techniques outperform ensembles and cost sensitive approaches.

Intrusion detection in computer networks is a important topic in information security. Due to numerous cases of security breaches that caused economic and social losses in recent years, this topic has been the subject of several studies in order to mitigate problems related to network intrusion and computer attacks. Information security systems have been using different techniques for network intrusion detection. However, with the development of communication mechanisms and consequently with the increase in data traffic, some techniques used for intrusion detection lost their information processing capability. The emergence of new forms of attacks on computer systems also contribute to the depreciation of some of the existing tools. In this scenario, new techniques capable of processing large amounts of information and that perform proactive discovery of new attack vectors are necessary. This paper presents a study on the use of a data stream mining technique known as Hoeffding Adaptive Tree to create a predictive model for network intrusion detection. The experiments performed in this work show the effectiveness of this technique when applied to a database of computer network attacks.

This paper proposes a classification approach in which monolithic and multiple classifier systems are combined in a cascading fashion. The rationale behind that is to deal with the existing trade-off between the need for increasing the accuracy, while reducing the complexity of the classification method. In other words, the idea is to offer an interesting strategy to conciliate the different levels of efforts necessary to deal with easy and hard patterns usually observed in a classification problem. The experimental results have shown that for some problems more than 90% of the instances can be processed in the first step of the cascade, saving efforts by avoiding the use of the second step in which a more complex classification method is used. It means that for some problems the reduction of the classification cost achieved more than 70% when compared to the use of an MCS. In addition to this interesting classification cost reduction, the cascade approach has shown to be able of improving the classification accuracy up to 15.19 percentage points.

The ever increasing data generation confronts both practitioners and researchers on handling massive and sequentially generated amounts of information, the so-called data streams. In this context, a lot of effort has been put on the extraction of useful patterns from streaming scenarios. Learning from data streams embeds a variety of problems, and by far, the most challenging is concept drift, i.e. changes in data distribution. In this paper, we focus on a specific type of drift uncommonly assessed in the literature: feature drifts. Feature drifts occur whenever a subset of features becomes, or ceases to be, relevant to the concept to be learned. We propose and review several feature drifting data stream generators and use them to benchmark state-of-the-art data stream classification algorithms and their combination with drift detectors. Results show that, although drift detectors enable slight quicker recovery to feature drifts, best results are obtained by Hoeffding Adaptive Tree, the only learner that performs dynamic feature selection as streams progress.

Extracting useful knowledge from data streams is problematic, mainly due to changes in their data distribution, a phenomenon named concept drift. Recently, studies have shown that most of existing algorithms for learning from data streams do not encompass techniques for a specific kind of drift: feature drifts. Feature drifts occur when features become, or cease to be, relevant to the learning task. In this paper, we propose an extension to the k-nearest neighbor classifier, so its distances' computations are weighted according to their current discriminative power. On our proposal, the discriminative power of features is given by entropy, which is swiftly computed over a sliding window. Empirical evidence shows that our approach is able to overcome several existing algorithms in accuracy and feature drift adaptation, while at the expense of bounded processing time and memory space.

Different dynamic classifier selection techniques have been proposed in the literature to determine among diverse classifiers available in a pool which should be used to classify a test instance. The individual competence of each classifier in the pool is usually evaluated taking into account its accuracy on the neighborhood of the test instance in a validation dataset. In this work we investigate the possible contribution of considering during the classifier evaluation the use of features related to the problem complexity. Since usually the pool generation technique does not assure diversity, the idea is to consider diversity during the selection. Basically, we select a classifier trained in subset of data showing similar complexity than that observed in neighborhood of the test instance. We expect that this similarity in terms of complexity allow us to select a more competent classifier. Experiments on 30 classification problems representing different levels of difficulty have shown that the proposed selection method is comparable to well known dynamic selection strategies. When compared with other DS approaches it was able to win on 123 over 150 experiments. This promising results indicate that further investigation must be done to increase diversity in terms of data complexity during the process of pool generation.

Trust and reputation mechanisms are part of the logical protection of intelligent agents, preventing malicious agents from acting egotistically or with the intention to damage others. Several studies in Psychology, Neurology and Anthropology claim that emotions are part of human's decision making process. However, there is a lack of understanding about how affective aspects, such as emotions, influence trust or reputation levels of intelligent agents when they are inserted into an information exchange environment, e.g. an evaluation system. In this paper we propose a reputation model that accounts for emotional bounds given by Ekman's basic emotions and inductive machine learning. Our proposal is evaluated by extracting emotions from texts provided by two online human-fed evaluation systems. Empirical results show significant agent's utility improvements with p <; .05 when compared to non-emotion-wise proposals, thus, showing the need for future research in this area.

Existing approaches to coalition formation are generally gross simplifications of real problems of resource allocation where experience, reputation, and time optimization should be considered, although they are not usually studied together. To overcome this issue, this study proposes a dynamic and distributed social coalition formation model, that reproduces real-world environments where interactions are ruled by an underlying network that adapts itself based on the best updated reputation of local neighbors, in order to bring together individuals better suited for efficient cooperation. In this environment, agents possessing different levels of expertise must be organized to provide the most advantageous partnerships for the purpose of solving tasks, and an execution order of task's subtasks is defined to favor the use and release of agents' resources. To achieve this objective, we based our proposal on a coalitional skill game (CSG) approach, which organizes the use of resources by time commitment, and calculates and exploits the temporal reputation of heterogeneous agents to improve the utility of coalitions. Our experiments with different initial social networks allowed us to evaluate the effectiveness of this proposal and provided elements to exploit the advantages of an optimized social structure in a connected world.

Service-Oriented Architecture (SOA) is a paradigm for software development that has been increasingly adopted for factory automation. In SOA, services are independently developed and a central engine orchestrates their functional behavior according to the process workflow. If on one hand this orchestration is required to maximize performance and productivity, i.e., the software is required to be maximally permissive, on the other hand, implementing a service orchestrator is a creative task which cannot be totally automated. Furthermore, industrial processes tend to be very large, making it difficult to empirically provide in-advance quality guarantees for industrial SOA-based applications. In this paper, we show how maximally permissive and deadlock-free service orchestrators can be implemented. We propose a model for each activity that compose a SOA programming language. Then, we show how pieces of a workflow can be individually represented by combining activity models. Afterwards, we specify the logical behavior of the workflow in order to organize those pieces and reproduce the orchestration effect. By using controllability concepts, we finally compute from the orchestrator a version of it that formally provides certain quality guarantees. Examples illustrate the approach.

Mining data streams is of the utmost importance due to its appearance in many real-world situations, such as: sensor networks, stock market analysis and computer networks intrusion detection systems. Data streams are, by definition, potentially unbounded sequences of data that arrive intermittently at rapid rates. Extracting useful knowledge from data streams embeds virtually all problems from conventional data mining with the addition of single-pass real-time processing within limited time and memory space. Additionally, due to its ephemeral nature, it is expected that streams undergo changes in its data distribution denominated concept drifts. In this work, we focus on one specific kind of concept drift that has not been extensively addressed in the literature, namely feature drift. A feature drift happens when changes occur in the set of features, such that a subset of features become, or cease to be, relevant to the learning problem. Specifically, changes in the relevance of features directly imply modifications in the decision boundary to be learned, thus the learner must detect and adapt to according to it. Timely detection and recover from feature drifts is a challenging task that can be modeled after a dynamic feature selection problem. In this paper we survey existing work on dynamic feature selection for data streams that acts either implicitly or explicitly. We conclude that there is a need for future research in this area, which we highlight as future research directions.

This article presents a multi-phase bidding model that is primarily for reverse bidding (1:N) and can also be used later for bilateral bidding (1:1). This multi-phase approach excels by competing for the lowest price and relativizes a subtle “lose-win” relationship of its own for the reverse auction through a second phase of negotiation. The latter is limited to a bilateral relationship and is applied, if necessary, between the purchaser and the second or third best offer of the reverse auction. Experiments were conducted with stationary or adaptive negotiation agents using learning techniques to conduct negotiation policy, using a genetic algorithm to characterize the opponent's preferences and configure the generation of interesting offers. The results showed the influence that an aggressive bidder has on the process as a whole and also what can be done to minimize this effect.

This paper presents a planning approach using Case-Based Reasoning (CBR) modeled as a Subsumption Architecture to generate plans for driving trains. The main idea of a planning strategy is to generate a sequence of actions for an agent, which can use these actions to change its environment. CBR allows using prior experiences for new task assignments. In the proposed ap-proach, each previous experience (if not applicable) is adjusted us-ing one or more adaptation methods like substitutive and genetic algorithm. Our interest is to create a flexible architecture for an agent and apply it to simulate train conductions. We expect that the plans generated by this approach generate better results com-pared to another studies already developed for the area mainly considering fuel consumption and travel time.

This work encompasses the development of a new ensemble classifier that uses a Social Network abstraction for Data Stream Classification, namely the Social Adaptive Ensemble (SAE). In the context of data stream classification, concept drift is considered one of the most difficult and important issues to be addressed. Ensemble classifiers can be successfully applied to data streams as long as the ensemble efficiently adapts itself in the occurrence of a concept drift. SAE algorithm inherits strategies from other ensemble methods, such as Online Bagging [4] and DWM [2], and merge these with the notion of connectivity between similar classifiers w.r.t. their individual predictions. The relational data obtained through measuring similarities between classifiers is used to arrange ensemble members in a social network structure that allows us to identify subgroups (subnetworks) of highly similar classifiers. Being able to identify similar classifiers allows us to implement a combination strategy that first combines predictions within similar classifiers and later combines these into the final prediction. Moreover, this combination strategy assigns more weight to recently added classifiers predictions during concept drifts, since these are dissimilar to all other existing classifiers. The similarity between classifiers is also used to identify and remove redundant classifiers. This effectively saves systems resources and sometimes improves accuracy. We present empirical experiments with synthetic data streams containing abrupt, gradual and no drift showing that SAE is a valid option for stream classification, especially when data stream characteristics (e.g. presence of abrupt drifts) are previously unknown and system resources, such as CPU time and memory space, are a concern.

This paper presents an Intelligent System, based on a dynamic time table definition, which coordinates the overtaking process of trains traveling in the same section of a railroad through a crossing loop. Each train involved on the overtaking process is represented by an intelligent agent capable of taken his actions based on his relative position on the railroad and his scheduling. The main goal of these agents is to react during the driving to avoid that more than one train stays on the same section of track at the same time (resource concurrency), and also to avoid unnecessary halts. The agents actions are previously defined, based on a simulation of the journey to the next crossing loop. For each stretch the agents selects another agent to be his coordinator, based on specific criteria. Then, each agent creates its action policy based on his local view and the data of the coordinator. The coordinator receives the actions and validates them generating a time table containing the actions for the next stretch, called dynamic time table. The definition of the action policy occurs dynamically, as each agent simulates the next step of the journey virtually and takes the decisions at runtime. The communication between the agents is done through the environment, where each agent updates his relative location and time. The main goal of the Intelligent System is the coordination of the trains focusing on reducing fuel consumption and also reducing the travel time. This is chased avoiding unnecessary halts, collisions of the trains and driving the trains with a Cruising Speed. The best simulations results achieved a 33.72% reduction in fuel consumption and a 33.30% reduction on travel time.

In this paper we propose an architecture of intelligent agent for automatic locomotives operating. The system agent generates its action policy using a set of resources, such as type of railway, composition, belief perception and reasoning about the actions. The focus of the operator agent is directed to the choice of acceleration points (gear) and preparation of travel plans in a journey guided by goals and objectives. The system is equipped with a module capable to plan the actions to move the vehicle from an initial point P to an end point Q and an executor module that implements the generated plan and modifies the state of the environment. For this purpose, we use the mental model that is based on the triple Belief, Desire and Intention (BDI) to which the perception of the agent is guaranteed by a set of sensors that provide speed information, position and breaks condition. The main focus on this research is the usage of mental model BDI for the resolution of a problem that combines travel naturally conflicting factors, such as safety, time and fuel consumption. Experimental results show that the developed architecture using the mental model BDI increases the efficiency of autonomous vehicles operating.

This paper presents a planning approach using Case-Based Reasoning (CBR) to generate plans for driving trains. The main idea of a planning strategy is to generate a sequence of actions for an agent, which can use these actions to change its environment. CBR allows using prior experiences in the situation assessment task. In the proposed approach, each previous experience (if not applicable) is adjusted resulting in cases specializations. Our interest is reducing the number of corrections triggered when a case retrieved is not applicable, based on these specializations. Experiments showed that the plans generated using this proposed method had a significant increase in the number of cases recovered satisfactorily, also reducing the need of adaptations for the cases recovered.

This paper presents a methodology to obtain rules of conduction from a set of data captured from sensors placed at a train as well data of actions executed by drivers. These actions result in a history H. The knowledge discovered is put in practice in a driving simulator and the result of the simulated actions generates a history H'. The validation of the discovered knowledge is done in an objective manner, which is calculated as a degree of similarity between the records. This degree of similarity reflects the performance of knowledge discovery process, which in experiments was around 85%. This degree of similarity represents how next were H and H.

This paper describes an intelligent approach based on agents that are able to drive and coordinate trains on stretches of railway line containing a crossing loop. Halts close to or even in crossing loops lead to increased consumption of fossil fuels, longer journey times and exhaustion of track capacity. In this paper the agents make use of a set of resources - railway line characteristics, train characteristics, driving rules and information about other trains - to generate their action policy. The agents perception is guaranteed by a set of sensors that provide data such as speed, position and information about the line. The tasks that the agent performs include carrying out actions such as increasing or reducing the speed of the train. The main objective of this study was to avoid unnecessary halts, which are the main cause of increased fuel consumption and journey time. Our results show that strong reductions can be made in terms of fuel consumption (average reduction of 25.5%), journey time (average reduction of 22.5%) and exhaustion of track capacity. Simulations were performed in which traditional driving techniques, with halts at several points along the stretch of track, were compared with driving performed by the multi-agent system, without any halts.

This paper presents an intelligent approach based on software agents capable of conducting and coordination trains in stretches of single railway track, aiming to reduce the utilization of railway and environment impacts. In the Brazilian rail modal, due to the low duplication of tracks, trains that journey on single railways should accomplish required halts, in order to wait for other trains to use the crossing loop safely. The technological evolution resulted on the appearance of new railway traffic system control. However, systems that rely on software agents are not well explored yet. Therefore, this paper elaborated a Multi-Agent System capable of simulating railway environment using agent drivers and agents with a highest level in managing the railway tracks. The behaviour of agents was based on specialized rules of conduction. The coordination between them occurs through message exchanges, always aiming to avoid halts during the journey. Results have shown an strong average reduction of 22.5% in journey time and 25.5% in fuel consumption when compared to journeys using the traditional method of conduction. The reduction, not only on the journey time, but also on the fuel consumption, entails on the decrease of CO2emission.

In supply chains of the petroleum industry, maintaining a balance between production and consumption of petroleum derivatives is a crucial issue. Basically, this kind of chain has several elements like producer bases, consumer bases, intermediary terminals that are linked by a transport network. These elements should cooperate to reach the global balance of the system with a minimum transportation cost. In this context, this paper proposes and compares two solutions based on auctions carried out by agents, which represent the elements of the aforementioned chain. These solutions are characterized respectively by the execution of sequential auctions and simultaneous auctions in order to consumers bid for batches of crude oil derivatives. In the comparative tests, the solution based on simultaneous auctions is better than the sequential one mainly because of the intense cooperation among agents.

This paper presents the development of an intelligent agent used to assist vehicle drivers. The agent has a set of resources to generate its action policy: road and vehicle features and a knowledge base containing conduct rules. The perception of the agent is ensured by a set of sensors, which provide the agent with data such as speed, position and conditions of the brakes. The main agent behaviour is to carry out action plans involving: increase, maintain or reduce speed. The main effort of this research was the induction of conduct rules from data of previous trips. These rules form a classifier used for the selection of actions forming the conduction plan. Results observed with the experiments have showed that the proposed classifier increases the efficiency throughout the conduction of vehicles.

We present a model for the certification of trust in multi-agent systems based on encryption. The objective is to raise the level of efficiency that client agents have when contracting specialized service agents. We make three hypotheses: (i) client agents are able to measure and inform the quality of a service they receive from a service agent; (ii) distributed certificate control is possible because every service agent stores the certificates it receives from its client agents and, (iii) the content of a certificate can be considered safe as long as the public and private keys used to encrypt the certificate remain safe. This approach reduces some weak points of trust models that rely on the direct interaction between service and client agents (direct trust) or those that rely on testimony obtained from client agents (propagated trust). Simulation showed that encrypted certificates of trust improved the efficiency of client agents when choosing their service provider agents. The reason seems to be that the reputation of a given service provider agent is based on the reputation it has among the totality of client agents that used its services.

This paper introduces a new agent-based algorithm for scheduling in CSCWD. Distributed artificial intelligence provides a lot of research areas, including CSCWD, with efficient decentralized optimization and problem solving techniques. In this paper we focus on how DCOP (distributed constraint optimization problem) can be used for scheduling in CSCWD, discussing a new algorithm. Our algorithm is almost-complete, providing the best solution most the times. However, it saves a lot of computational resources, being useful in situations where other state of the art algorithms are not feasible. The results are quite encouraging, showing that our algorithm outperforms easily two well-known DCOP algorithms in terms of runtime, number of messages, size of messages and throughput.

This paper proposes an approach to generate offers and counter-offers in a bilateral negotiation process between cognitive agents with the learning machine capacities. The approach is configured, as each participant of a negotiation process improve their satisfaction degree, gaining knowledge from prior experience, i.e., in the next negotiation session, each participant can individually use this knowledge to redefine the configuration parameters of strategies and tactics to generate offers and counter-offers. Each agent refines their strategies using a genetic algorithm application based on the historical offers and counter-offer exchanges dataset. This context can lead to a new dimension in CSCW system development. This approach was tested in a simulated bilateral negotiation environment, which involved, for example, a buyer agent and a seller agent. The discussions about the results confront different negotiation sessions comparing agents provided with strategy and tactics reconfiguration capabilities and agents using static strategies and tactics.

This paper presents a novel method for the classification of images that combines information extracted from the images and contextual information. The main hypothesis is that contextual information related to an image can contribute in the image classification process. First, independent classifiers are designed to deal with images and text. From the images color, shape and texture features are extracted. These features are used with a neural network (NN) classifier to carry out image classification. On the other hand, contextual information is processed and used with a Naive Bayes (NB) classifier. At the end, the outputs of both classifiers are combined through heuristic rules. Experimental results on a database of more than 5,000 HTML documents have shown that the combination of classifiers provides a meaningful improvement (about 16%) in the correct image classification rate relative to the results provided by the NN classifier alone.

A common problem in organizations is to identify people with the right competencies to form a specialized team, in academic or industrial environments, which is capable of executing a self-managed project of Research and Development (R&D). This work presents a technique that allows for identifying people who have the most appropriated competencies to form a R&D team extracting information from their curriculum vitas (CVs). Information extraction in this work is carried out by means of textual retrieval techniques in document databases. The system was evaluated with data of real projects producing expressive results when identifying people to participate in research projects.

This paper aims to conceive a system capable of assessing the understanding an individual possesses on an arbitrary subject. A hard, essential question regarding a collaborative environment is how to evaluate the knowledge an individual maintains about a specific domain. Such evaluation is usually accomplished with collective works or examinations applied individually. However, techniques such as those are subjected to idiosyncratic factors related to both the tutor (level of experience, mood, affinity, etc.) and the team member. In this paper, we used some techniques from the field of natural language understanding based on concepts like dynamic memory, case-based reasoning and semantic parser. We discuss experiments, carried out within one such educational environment, which turned out to be coherent according to the opinion of the tutors involved

In this paper we present a new meta-learning approach for Knowledge Integration. To generate accurate classifiers one can use combination techniques like Stacking, Bagging and Boosting. Such techniques are used for the generation of vote committees that produce decisions much more accurate than simple base classifiers. It is known that even using quite small partitions of the training database such techniques produce much more accurate decisions than a simple base classifier that uses all the training data. This is suitable for solving scalability problems. However, such techniques can not learn understandable knowledge, what is a drawback from the Knowledge Discover process point-of-view. To solve these problems, we introduce in this paper a Knowledge Integration technique capable of generate accurate and understandable rule sets taking as input base classifiers generated by a rule induction algorithm. Such rule sets are combined into a single rule set that, when evaluated over test instances, presents a better accuracy than any individual rule set and often outperforms Bagging and AdaBoosting.

This article concerns a particular aspect of knowledge management (KM) in design teams: the awareness. It consists in keeping the members of a team conscious about the activities of their colleagues. We present a mechanism to determine the interest center of a user built from the traces of his/her computer operations (e.g. documents, Web pages). Each user is notified of similar interest centers thanks to a society of agents. We discuss in this article the construction of the centers of interest and the notification protocol.

This paper describes a personal assistant (PA) for knowledge management. Groupware and collaborative software do not explore domain knowledge efficiently. This information is important to organize the information and to help the user in specialized work. Here we describe a PA using domain knowledge to help a user in a collaborative environment. The PA records user's tasks uses domain knowledge to organize them, and offers help and propositions regarding the user's work. The PAs communicate with others PAs to exchange users' information. Therefore, one can reuse experiences of others users. The interaction with the user occurs through window dialogs and by means of natural language. PAs are used in an educational environment to aid remote groups of students to do collaborative mechanical engineering design. We make a brief review about PAs and then introduce our approach.