Browsing by Author "Oliveira, Thays Aparecida de"

Now showing 1 - 5 of 5
  • No Thumbnail Available
    Item
    5th International Conference on Variable Neighborhood Search (ICVNS'17).
    (2018) Coelho, Vitor Nazário; Santos, Haroldo Gambini; Coelho, Igor Machado; Penna, Puca Huachi Vaz; Oliveira, Thays Aparecida de; Souza, Marcone Jamilson Freitas; Sifaleras, Angelo
    This volume presents selected, peer-reviewed, short papers that were accepted for presentation in the 5th International Conference on Variable Neighborhood Search (ICVNS'17) which was held in Ouro Preto, Brazil, during October 2–4, 2017.
  • No Thumbnail Available
    Item
    EEG time series learning and classification using a hybrid forecasting model calibrated with GVNS.
    (2017) Coelho, Vitor Nazário; Coelho, Igor Machado; Coelho, Bruno Nazário; Souza, Marcone Jamilson Freitas; Guimarães, Frederico Gadelha; Luz, Eduardo José da Silva; Barbosa, Alexandre Costa; Coelho, M. N.; Netto, Guilherme Gaigher; Pinto, Alysson Alves; Elias, M. E. V.; G. Filho, D. C. O.; Oliveira, Thays Aparecida de
    Brain activity can be seen as a time series, in particular, electroencephalogram (EEG) can measure it over a specific time period. In this regard, brain fingerprinting can be subjected to be learned by machine learning techniques. These models have been advocated as EEG-based biometric systems. In this study, we apply a recent Hybrid Focasting Model, which calibrates its if-then fuzzy rules with a hybrid GVNS metaheuristic algorithm, in order to learn those patterns. Due to the stochasticity of the VNS procedure, models with different characteristics can be generated for each individual. Some EEG recordings from 109 volunteers, measured using a 64-channels EEGs, with 160 HZ of sampling rate, are used as cases of study. Different forecasting models are calibrated with the GVNS and used for the classification purpose. New rules for classifying the individuals using forecasting models are introduced. Computational results indicate that the proposed strategy can be improved and embedded in the future biometric systems.
  • No Thumbnail Available
    Item
    Generic Pareto local search metaheuristic for optimization of targeted offers in a bi-objective direct marketing campaign.
    (2016) Coelho, Vitor Nazário; Oliveira, Thays Aparecida de; Coelho, Igor Machado; Coelho, Bruno Nazário; Fleming, Peter J.; Guimarães, Frederico Gadelha; Ramalhinho, Helena; Souza, Marcone Jamilson Freitas; Talbi, El-Ghazali; Lust, Thibaut
    Cross-selling campaigns seek to offer the right products to the set of customers with the goal of maximizing expected profit, while, at the same time, respecting the purchasing constraints set by investors. In this context, a bi-objective version of this NP-Hard problem is approached in this paper, aiming at maximizing both the promotion campaign total profit and the risk-adjusted return, which is estimated with the reward-to-variability ratio known as Sharpe ratio. Given the combinatorial nature of the problem and the large volume of data, heuristic methods are the most common used techniques. A Greedy Randomized Neighborhood Structure is also designed, including the characteristics of a neighborhood exploration strategy together with a Greedy Randomized Constructive technique, which is embedded in a multi-objective local search metaheuristic. The latter combines the power of neighborhood exploration by using a Pareto Local Search with Variable Neighborhood Search. Sets of non-dominated solutions obtained by the proposed method are described and analyzed for a number of problem instances.
  • No Thumbnail Available
    Item
    A hybrid variable neighborhood search algorithm for targeted offers in direct marketing.
    (2015) Oliveira, Thays Aparecida de; Coelho, Vitor Nazário; Souza, Marcone Jamilson Freitas; Boava, Diego Luiz Teixeira; Boava, Fernanda Maria Felício Macêdo; Coelho, Igor Machado; Coelho, Bruno Nazário
    This paper focuses on the targeted offers problem in direct marketing campaigns. The main objective is to maximize the feedback of customers purchases, offering products for the set of customers with the highest probability of positively accepting the offer and, at the same time, minimizing the operational costs of the campaign. Given the combinatorial nature of the problem and the large volume of data, involving instances with up to one million customers, approaches solely based on mathematical programming methods, said exact, appear limited and infeasible. In this paper, the use of a hybrid heuristic algorithm, based on the Greedy Randomized Adaptive Search Procedures and General Variable Neighborhood Search, is proposed. Computational experiments performed on a set of test problems from the literature show that the proposed algorithm was able to produce competitive solutions.
  • No Thumbnail Available
    Item
    A VNS approach for book marketing campaigns generated with quasi-bicliques probabilities.
    (2017) Oliveira, Thays Aparecida de; Coelho, Vitor Nazário; Ramalhinho, Helena; Souza, Marcone Jamilson Freitas; Coelho, Bruno Nazário; Rezende, Daniel C.; Coelho, Igor Machado
    This paper focuses on Book Marketing Campaigns, where the benefit of offering each book is calculated based on a bipartite graph (biclique). A quasi Biclique problem is assessed for obtaining the probabilities of success of a given client buy a given book, considering it had received another book as free offer. The remaining optimization decision problem can be solved following the Targeted Offers Problem in Direct Marketing Campaigns. The main objective is to maximize the feedback of customers purchases, offering books to the set of customers with the highest probability of buying others ones from its biclique and, at the same time, minimizing campaign operational costs. Given the combinatorial nature of the problem and the large volume of data, which can involve real cases with up to one million customers, metaheuristics procedures have been used as an efficient way for solving it. Here, a hybrid trajectory search based algorithm, namely GGVNS, which combines the Greedy Randomized Adaptive Search Procedures and General Variable Neighborhood Search, is used. The strategy for generating the quasi Biclique problem is described and a new instance generator for the TOPDMC is introduced. Computational results regarding the GGVNS algorithm shows it is able to find useful and profitable sets of clients.