Hendrik T. Macedo

Users' satisfaction in recommendation systems for groups: an approach based on noncooperative games

A major difficulty in a recommendation system for groups is to use a group aggregation strategy to ensure, among other things, the maximization of the average satisfaction of group members. This paper presents an approach based on the theory of noncooperative games to solve this problem. While group members can be seen as game players, the items for potential recommendation for the group comprise the set of possible actions. Achieving group satisfaction as a whole becomes, then, a problem of finding the Nash equilibrium. Experiments with a MovieLens dataset and a function of arithmetic mean to compute the prediction of group satisfaction for the generated recommendation have shown statistically significant results when compared to state-of-the-art aggregation strategies, in particular, when evaluation among group members are more heterogeneous. The feasibility of this unique approach is shown by the development of an application for Facebook, which recommends movies to groups of friends.

Generation of coalition structures to provide proper groups' formation in group recommender systems

Group recommender systems usually provide recommendations to a fixed and predetermined set of members. In some situations, however, there is a set of people (N) that should be organized into smaller and cohesive groups, so it is possible to provide more effective recommendations to each of them. This is not a trivial task. In this paper we propose an innovative approach for grouping people within the recommendation problem context. The problem is modeled as a coalitional game from Game Theory. The goal is to group people into exhaustive and disjoint coalitions so as to maximize the social welfare function of the group. The optimal coalition structure is that with highest summation over all social welfare values. Similarities between recommendation system users are used to define the social welfare function. We compare our approach with K-Means clustering for a dataset from Movielens. Results have shown that the proposed approach performs better than K-Means for both average group satisfaction and Davies-Bouldin index metrics when the number of coalitions found is not greater than 4 (K <= 4) for a population size of 12 (N = 12).

Grouping Similar Trajectories for Carpooling Purposes

Vehicle congestion is a serious concern in metropolitan areas. Some policies have been adopted in order to soften the problem: construction of alternative routes, encouragement for the use of bicycles, improvement on public transportation, among others. A practice that might help is carpooling. Carpooling consists in sharing private vehicle space among people with similar trajectories. Although there exist some software initiatives to facilitate the carpooling practice, none of them actually provides some key facilities such as searching for people with similar trajectories. The way in which such a trajectory is represented is also central. In the specific context of carpooling, the use of Points of Interest (POI) as a method for trajectory discretization is rather relevant. In this paper, we consider that and other assumptions to propose an innovative approach to generate trajectory clusters for carpooling purposes, based on Optics algorithm. We also propose a new similarity measure for trajectories. Two experiments have been performed in order to prove the feasibility of the approach. Furthermore, we compare our approach with K-means and Optics. Results have showed that the proposed approach has results similar for Davies-Boulding index (DBI).

A social network for carpooling

The larger amount of motor vehicles produces many problems to cities around the world as poor urban mobility, environment concerns and peoples health. The encouragement to use alternative means of transport, such as bicycles or subways is one of the steps to solve the problem. Another way is to encouraging the optimal usage of private automobiles. In Brazil, for instance, several studies show an underused space in cars and great potential for the use of carpooling and carsharing. Unfortunately, there is not such a unified technological mean to properly support these alternatives. In this paper we propose and implement a social network to support the use of carpooling. We show that there is a favorable context for the use of this social network with the specific purpose to share rides among users. We describe the architecture of the system and implementation details. We present results about systemss accessibility and its acceptability by people around the globe.

Vehicle driving analysis in regards to fuel consumption using Fuzzy Logic and OBD-II devices

Smart City is a concept used to describe a city that uses technology in order to improve the quality of life in the urban environment. One way to contribute to the improvement of urban life is the optimized use of resources. Given the high price of gasoline in Brazil and aiming to improve the quality of urban life, we developed a mobile application that uses fuzzy logic and data read from a OBD-II Bluetooth device (ELM 327) to classify the driver in regards to the quality of vehicle conduction, in order to encourage fuel consumption reduction. The set of data to be read was defined by expert knowledge. Results show that is possible to estimate instant consumption. Experiments carried out in routine traffic situation in Aracaju/SE city have shown an accuracy of more than 85% if compared to the onboard computer information.

Flood monitoring in smart cities based on fuzzy logic about urban open data

A new method for urban flood monitoring based on fuzzy logic for issue alerts on a geographic map was showed. The model is designed to receive information from remote sensing water level over TCP/IP protocol and operate intelligently based on fuzzy rules with inference by Mamdani model from urban open data, providing free information access containing flood alert states and flooding with georeferencing that can help cities to become smart cities. The model was validated by case studies analysis with real sensors that generate a set of pre-planned test data. The issue alerts on geographical maps in real-time corroborated presented in this work for the acceptance of fuzzy logic in the urban monitoring flood with high performance, being able to act autonomously providing information to drivers for avoid undesirable routes.

Multi-kernel approach to Parallelization of EM Algorithm for GMM Training

Most machine learning algorithms need to handle large datasets. This feature often leads to limitations on processing time and memory. The Expectation-Maximization (EM) is one of such algorithms, which is used to train one of the most commonly used parametric statistical models, the Gaussian Mixture Models (GMM). All steps of the algorithm are potentially parallelizable once they iterate over the entire data set. In this work, we propose a parallel implementation of EM for training GMM using CUDA cores. Experimentation scenario consists of five different datasets and four metrics. Results show a speedup of 12.7 if compared to sequential version. With coalesced access to CUDA global memory and shared memory usage, we have achieved up to 99.4% of actual occupancy, regardless the number of Gaussians considered.

Automatic code generation of SIMUROSOT game strategies: an approach based on finite state machines

Robot soccer is a multidisciplinary research field of growing academic interest. Scientific competitions worldwide have driven major technological advances in the field, since research institutes and universities often enroll their robotic teams and present the developments achieved during the previous year. The FIRA is a leading international organization that promotes such type of competition. One of its various competition categories is that of virtual robots competing in a simulation platform called SIMUROSOT. This 3D simulation platform is available so that researchers and students can develop robot teams and test their performance appropriately and repeatedly. The major problem with the mainstream encoding mechanism used in the preparation of robot teams for SimuroSot is the lack of a formal process to standardize the programming and model the game strategies. Such lack of standardization causes several limitations on the competition as a science promoter, namely: (1) the difficulty of understanding others code, (2) absence of abstraction for the conceiving of game strategies, (3) need for high programming skills in order to produce effective game strategies and (4) lack of clear separation between robots movimentation primitives and specific application domain concerns - in this case, the soccer domain. This paper proposes a semi-formal specification for the modeling of team game strategies to the FIRA SIMUROSOT category. This specification defines two levels of mapping: (1) mapping of the application domain information - in this case, the soccer competition - to Finite State Machines and (2) mapping of Finite State Machines that represent game strategies of a team to their respective coding in a target programming language, making use of coding standards that ensure the readability and maintainability of the final code. Encodings of game strategies in the programming language Lingo has been provided following the proposal. The error-free execution of such encodings on the simulation platform shows the proposal feasibility.

Automated feeding of POI base for the generation of route descriptions

Mashups is a new way to develop Web applications by combining preexistent Web services, features and content in order to provide new features and services. Location systems are maybe the most popular mashups. An example of a location system is RotaCerta. This system makes use of Google Maps API to generate textual route descriptions between two locations within a city. RotaCerta novelty is the use of Points of Interest (POI) in the generated descriptions so to facilitate the route comprehension and assimilation by the user. RotaCerta suffers, however, from a severe limitation on the need for manual feeding of the POIs base for each city of interest. This work is exhausting, costly, and greatly limits its usage. This paper proposes a mechanism for automated feeding of POI base for RotaCerta system by means of Google Places. Experiments were performed for three different cities in Brazil and the results confirm the gain of intuitiveness in the textual description of the generated paths and demonstrated the elimination of previous limitation and the consequent encouragement of its large scale usage.

Improving CUDA™ C/C++ encoding readability to foster parallel application development

Graphical Processing Units (GPUs) have recently been used to enable parallel application development. The most prominent initiative has been provided by NVIDIA™ with the so-called CUDA™ architecture, designed to GeForce™ graphic cards. However, even with CUDA C-like programming language, parallel codification remains somewhat awkward if compared to sequential codification. The programmer still has to deal with low-level hardware details such as generation and synchronization of threads and GPU tracks and sectors. In this paper, we propose a programmer-friendly interface for CUDA-C programming, in such a way that most hardware details are hidden from the programmer. We show how code readability is improved without undermining parallel execution performance.